refactor: deep Codex v4 API conversion for all 20 team skills

Upgrade all team-* skills from mechanical v3→v4 API renames to deep
v4 tool integration with skill-adaptive patterns:

- list_agents: health checks in handleResume, cleanup verification in
  handleComplete, added to allowed-tools and coordinator toolbox
- Named targeting: task_name uses task-id (e.g. EXPLORE-001) instead
  of generic <role>-worker, enabling send_message/assign_task by name
- Message semantics: send_message for supplementary cross-agent context
  vs assign_task for triggering work, with skill-specific examples
- Model selection: per-role reasoning_effort guidance matching each
  skill's actual roles (not generic boilerplate)
- timeout_ms: added to all wait_agent calls, timed_out handling in
  all 18 monitor.md files
- Skill-adaptive v4 sections: ultra-analyze N-parallel coordination,
  lifecycle-v4 supervisor assign_task/send_message distinction,
  brainstorm ideator parallel patterns, iterdev generator-critic loops,
  frontend-debug iterative debug assign_task, perf-opt benchmark
  context sharing, executor lightweight trimmed v4, etc.

60 files changed across 20 team skills (SKILL.md, monitor.md, role.md)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

.codex/skills/team-arch-opt/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: team-arch-opt
 description: Unified team skill for architecture optimization. Uses team-worker agent architecture with role directories for domain logic. Coordinator orchestrates pipeline, workers are team-worker agents. Triggers on "team arch-opt".
-allowed-tools: spawn_agent(*), wait_agent(*), send_input(*), close_agent(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*), mcp__ace-tool__search_context(*)
+allowed-tools: spawn_agent(*), wait_agent(*), send_message(*), assign_task(*), close_agent(*), list_agents(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*), mcp__ace-tool__search_context(*)
 ---
 
 # Team Architecture Optimization
@@ -54,7 +54,8 @@ Before calling ANY tool, apply this check:
 
 | Tool Call | Verdict | Reason |
 |-----------|---------|--------|
-| `spawn_agent`, `wait_agent`, `close_agent`, `send_input` | ALLOWED | Orchestration |
+| `spawn_agent`, `wait_agent`, `close_agent`, `send_message`, `assign_task` | ALLOWED | Orchestration |
+| `list_agents` | ALLOWED | Agent health check |
 | `request_user_input` | ALLOWED | User interaction |
 | `mcp__ccw-tools__team_msg` | ALLOWED | Message bus |
 | `Read/Write` on `.workflow/.team/` files | ALLOWED | Session state |
@@ -84,6 +85,8 @@ Coordinator spawns workers using this template:
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
   items: [
     { type: "text", text: `## Role Assignment
 role: <role>
@@ -107,11 +110,34 @@ pipeline_phase: <pipeline-phase>` },
 })
 ```
 
-After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ id })` each worker.
+After spawning, use `wait_agent({ targets: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ target: <name> })` each worker.
 
 **Inner Loop roles** (refactorer): Set `inner_loop: true`.
 **Single-task roles** (analyzer, designer, validator, reviewer): Set `inner_loop: false`.
 
+### Model Selection Guide
+
+Architecture optimization is reasoning-intensive. All analysis and design roles need high reasoning effort.
+
+| Role | reasoning_effort | Rationale |
+|------|-------------------|-----------|
+| analyzer | high | Deep structural analysis of codebase architecture |
+| designer | high | Architecture redesign requires careful reasoning about tradeoffs |
+| refactorer | high | Code transformations must preserve correctness |
+| validator | high | Validation must thoroughly check refactoring correctness |
+| reviewer | high | Code quality review demands deep understanding |
+
+Override in spawn_agent when needed:
+```
+spawn_agent({
+  agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
+  reasoning_effort: "high",
+  items: [...]
+})
+```
+
 ## User Commands
 
 | Command | Action |
@@ -161,6 +187,47 @@ After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect
 
 - [specs/pipelines.md](specs/pipelines.md) — Pipeline definitions, task registry, parallel modes
 
+## v4 Agent Coordination
+
+### Message Semantics
+
+| Intent | API | Example |
+|--------|-----|---------|
+| Queue supplementary info (don't interrupt) | `send_message` | Send codebase patterns to running analyzer |
+| Assign new work / trigger processing | `assign_task` | Assign fix task to refactorer after review feedback |
+| Check running agents | `list_agents` | Verify agent health during resume |
+
+### Agent Health Check
+
+Use `list_agents({})` in handleResume and handleComplete:
+
+```
+// Reconcile session state with actual running agents
+const running = list_agents({})
+// Compare with session.json active tasks
+// Reset orphaned tasks (in_progress but agent gone) to pending
+```
+
+### Named Agent Targeting
+
+Workers are spawned with `task_name: "<task-id>"` enabling direct addressing:
+- `send_message({ target: "ANALYZE-001", items: [...] })` -- queue analysis context without interrupting
+- `assign_task({ target: "REFACTOR-001", items: [...] })` -- assign fix task after review feedback
+- `close_agent({ target: "VALIDATE-001" })` -- cleanup by name
+
+### Merged Exploration Pattern
+
+For architecture analysis, analyzer may need broad codebase exploration. Consider spawning analyzer with `fork_context: true` when deep structural analysis of interconnected modules is needed:
+```
+spawn_agent({
+  agent_type: "team_worker",
+  task_name: "ANALYZE-001",
+  fork_context: true,   // Share coordinator's codebase context
+  reasoning_effort: "high",
+  items: [...]
+})
+```
+
 ## Error Handling
 
 | Scenario | Resolution |

.codex/skills/team-arch-opt/roles/coordinator/commands/monitor.md

@@ -41,6 +41,16 @@ Fan-out mode adds per-branch grouping. Independent mode adds per-pipeline groupi
 
 ## handleResume
 
+**Agent Health Check** (v4):
+```
+// Verify actual running agents match session state
+const runningAgents = list_agents({})
+// For each active_agent in tasks.json:
+//   - If agent NOT in runningAgents -> agent crashed
+//   - Reset that task to pending, remove from active_agents
+// This prevents stale agent references from blocking the pipeline
+```
+
 1. Read tasks.json, check active_agents
 2. No active agents -> handleSpawnNext
 3. Has active agents -> check each status
@@ -74,6 +84,7 @@ state.tasks[taskId].status = 'in_progress'
 // 2) Spawn worker
 const agentId = spawn_agent({
   agent_type: "team_worker",
+  task_name: taskId,  // e.g., "DESIGN-001" — enables named targeting
   items: [
     { type: "text", text: `## Role Assignment
 role: ${task.role}
@@ -109,20 +120,43 @@ state.active_agents[taskId] = { agentId, role: task.role, started_at: now }
 | Fan-out (REFACTOR-B{NN} done) | VALIDATE + REVIEW ready | Spawn both for that branch in parallel |
 | Independent | Any unblocked task | Spawn all ready tasks across all pipelines in parallel |
 
+**Cross-Agent Supplementary Context** (v4):
+
+When spawning workers in a later pipeline phase, send upstream results as supplementary context to already-running workers:
+
+```
+// Example: Send design analysis results to running refactorers
+send_message({
+  target: "<running-agent-task-name>",
+  items: [{ type: "text", text: `## Supplementary Context\n${upstreamFindings}` }]
+})
+// Note: send_message queues info without interrupting the agent's current work
+```
+
+Use `send_message` (not `assign_task`) for supplementary info that enriches but doesn't redirect the agent's current task.
+
 ### Wait and Process Results
 
 After spawning all ready tasks:
 
 ```javascript
-// 4) Batch wait for all spawned workers
-const agentIds = Object.values(state.active_agents).map(a => a.agentId)
-wait_agent({ ids: agentIds, timeout_ms: 900000 })
-
-// 5) Collect results
-for (const [taskId, agent] of Object.entries(state.active_agents)) {
-  state.tasks[taskId].status = 'completed'
-  close_agent({ id: agent.agentId })
-  delete state.active_agents[taskId]
+// 4) Batch wait — use task_name for stable targeting (v4)
+const taskNames = Object.keys(state.active_agents)
+const waitResult = wait_agent({ targets: taskNames, timeout_ms: 900000 })
+if (waitResult.timed_out) {
+  // Mark timed-out agents, close them, report to user
+  for (const taskId of taskNames) {
+    state.tasks[taskId].status = 'timed_out'
+    close_agent({ target: taskId })
+    delete state.active_agents[taskId]
+  }
+} else {
+  // 5) Collect results
+  for (const [taskId, agent] of Object.entries(state.active_agents)) {
+    state.tasks[taskId].status = 'completed'
+    close_agent({ target: taskId })  // Use task_name, not agentId
+    delete state.active_agents[taskId]
+  }
 }
 ```
 
@@ -159,6 +193,14 @@ Fix cycle tracking per branch in tasks.json `fix_cycles`:
 
 ## handleComplete
 
+**Cleanup Verification** (v4):
+```
+// Verify all agents are properly closed
+const remaining = list_agents({})
+// If any team agents still running -> close_agent each
+// Ensures clean session shutdown
+```
+
 Pipeline done. Generate report and completion action.
 
 Completion check by mode:

.codex/skills/team-arch-opt/roles/coordinator/role.md

@@ -6,7 +6,7 @@ Orchestrate team-arch-opt: analyze -> dispatch -> spawn -> monitor -> report.
 
 **You are a dispatcher, not a doer.** Your ONLY outputs are:
 - Session state files (`.workflow/.team/` directory)
-- `spawn_agent` / `wait_agent` / `close_agent` / `send_input` calls
+- `spawn_agent` / `wait_agent` / `close_agent` / `send_message` / `assign_task` calls
 - Status reports to the user / `request_user_input` prompts
 
 **FORBIDDEN** (even if the task seems trivial):
@@ -34,6 +34,8 @@ WRONG: Bash running build/test/lint commands          — worker work
 - Handle review-fix cycles with max 3 iterations
 - Execute completion action in Phase 5
 - **Always proceed through full Phase 1-5 workflow, never skip to direct execution**
+- Use `send_message` for supplementary context (non-interrupting) and `assign_task` for triggering new work
+- Use `list_agents` for session resume health checks and cleanup verification
 
 ### MUST NOT
 - Implement domain logic (analyzing, refactoring, reviewing) -- workers handle this
@@ -173,6 +175,16 @@ Delegate to @commands/monitor.md#handleSpawnNext:
    - auto_archive -> Archive & Clean (rm -rf session folder)
    - auto_keep -> Keep Active (status=paused)
 
+## v4 Coordination Patterns
+
+### Message Semantics
+- **send_message**: Queue supplementary info to a running agent. Does NOT interrupt current processing. Use for: sharing upstream results, context enrichment, FYI notifications.
+- **assign_task**: Assign new work and trigger processing. Use for: waking idle agents, redirecting work, requesting new output.
+
+### Agent Lifecycle Management
+- **list_agents({})**: Returns all running agents. Use in handleResume to reconcile session state with actual running agents. Use in handleComplete to verify clean shutdown.
+- **Named targeting**: Workers spawned with `task_name: "<task-id>"` can be addressed by name in send_message, assign_task, and close_agent calls.
+
 ## Error Handling
 
 | Error | Resolution |

.codex/skills/team-brainstorm/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: team-brainstorm
 description: Unified team skill for brainstorming team. Uses team-worker agent architecture with role directories for domain logic. Coordinator orchestrates pipeline, workers are team-worker agents. Triggers on "team brainstorm".
-allowed-tools: spawn_agent(*), wait_agent(*), send_input(*), close_agent(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*)
+allowed-tools: spawn_agent(*), wait_agent(*), send_message(*), assign_task(*), close_agent(*), list_agents(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*)
 ---
 
 # Team Brainstorm
@@ -53,7 +53,8 @@ Before calling ANY tool, apply this check:
 
 | Tool Call | Verdict | Reason |
 |-----------|---------|--------|
-| `spawn_agent`, `wait_agent`, `close_agent`, `send_input` | ALLOWED | Orchestration |
+| `spawn_agent`, `wait_agent`, `close_agent`, `send_message`, `assign_task` | ALLOWED | Orchestration |
+| `list_agents` | ALLOWED | Agent health check |
 | `request_user_input` | ALLOWED | User interaction |
 | `mcp__ccw-tools__team_msg` | ALLOWED | Message bus |
 | `Read/Write` on `.workflow/.team/` files | ALLOWED | Session state |
@@ -83,6 +84,8 @@ Coordinator spawns workers using this template:
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
   items: [
     { type: "text", text: `## Role Assignment
 role: <role>
@@ -106,7 +109,7 @@ pipeline_phase: <pipeline-phase>` },
 })
 ```
 
-After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ id })` each worker.
+After spawning, use `wait_agent({ targets: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ target: <name> })` each worker.
 
 **Parallel ideator spawn** (Full pipeline with N angles):
 
@@ -115,6 +118,8 @@ When Full pipeline has N parallel IDEA tasks, spawn N distinct team-worker agent
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: "ideator-<N>",
+  fork_context: false,
   items: [
     { type: "text", text: `## Role Assignment
 role: ideator
@@ -139,7 +144,29 @@ pipeline_phase: <pipeline-phase>` },
 })
 ```
 
-After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ id })` each worker.
+After spawning, use `wait_agent({ targets: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ target: <name> })` each worker.
+
+
+### Model Selection Guide
+
+| Role | model | reasoning_effort | Rationale |
+|------|-------|-------------------|-----------|
+| Ideator (IDEA-*) | (default) | high | Creative divergent thinking requires full reasoning |
+| Challenger (CHALLENGE-*) | (default) | high | Critical analysis of ideas needs deep reasoning |
+| Synthesizer (SYNTH-*) | (default) | medium | Aggregation and convergence over generation |
+| Evaluator (EVAL-*) | (default) | medium | Scoring and ranking against criteria |
+
+Override model/reasoning_effort in spawn_agent when cost optimization is needed:
+```
+spawn_agent({
+  agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
+  model: "<model-override>",
+  reasoning_effort: "<effort-level>",
+  items: [...]
+})
+```
 
 ## User Commands
 
@@ -178,6 +205,47 @@ After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect
 
 - [specs/pipelines.md](specs/pipelines.md) — Pipeline definitions and task registry
 
+## v4 Agent Coordination
+
+### Message Semantics
+
+| Intent | API | Example |
+|--------|-----|---------|
+| Share idea context across ideators | `send_message` | Send seed ideas to running ideator for cross-pollination |
+| Not used in this skill | `assign_task` | No resident agents -- all workers are one-shot |
+| Check running agents | `list_agents` | Verify parallel ideator health during Full pipeline |
+
+### Parallel Ideator Coordination (Full Pipeline)
+
+Full pipeline spawns N parallel ideators for different brainstorming angles. Use batch spawn + wait:
+
+```
+// Spawn N ideators in parallel, each with a different angle
+const ideatorNames = ["IDEA-001", "IDEA-002", "IDEA-003"]
+for (const name of ideatorNames) {
+  spawn_agent({ agent_type: "team_worker", task_name: name, fork_context: false, ... })
+}
+wait_agent({ targets: ideatorNames, timeout_ms: 900000 })
+// Collect all idea outputs, feed to challenger as upstream context
+```
+
+### Agent Health Check
+
+Use `list_agents({})` in handleResume and handleComplete:
+
+```
+// Reconcile session state with actual running agents
+const running = list_agents({})
+// Compare with tasks.json active_agents
+// Reset orphaned tasks (in_progress but agent gone) to pending
+```
+
+### Named Agent Targeting
+
+Workers are spawned with `task_name: "<task-id>"` enabling direct addressing:
+- `send_message({ target: "IDEA-002", items: [...] })` -- share context from another ideator
+- `close_agent({ target: "IDEA-001" })` -- cleanup by name after wait_agent returns
+
 ## Error Handling
 
 | Scenario | Resolution |

.codex/skills/team-brainstorm/roles/coordinator/commands/monitor.md

@@ -51,7 +51,7 @@ Worker completed. Process and advance.
    - Log team_msg with type "gc_loop_trigger" or "task_unblocked"
    - If skipping GC tasks, mark them as completed (skip)
 
-5. Close completed agent: `close_agent({ id: <agentId> })`
+5. Close completed agent: `close_agent({ target: <agentId> })`
 6. Proceed to handleSpawnNext
 
 ## handleCheck
@@ -69,6 +69,16 @@ Read-only status report, then STOP.
 
 ## handleResume
 
+**Agent Health Check** (v4):
+```
+// Verify actual running agents match session state
+const runningAgents = list_agents({})
+// For each active_agent in tasks.json:
+//   - If agent NOT in runningAgents -> agent crashed
+//   - Reset that task to pending, remove from active_agents
+// This prevents stale agent references from blocking the pipeline
+```
+
 1. Audit task list: Tasks stuck in "in_progress" -> reset to "pending"
 2. Proceed to handleSpawnNext
 
@@ -86,6 +96,7 @@ Find ready tasks, spawn workers, STOP.
       ```
       const agentId = spawn_agent({
         agent_type: "team_worker",
+        task_name: taskId,  // e.g., "IDEA-001" — enables named targeting
         items: [{ type: "text", text: `## Role Assignment
       role: <role>
       role_spec: ~  or <project>/.codex/skills/team-brainstorm/roles/<role>/role.md
@@ -99,10 +110,10 @@ Find ready tasks, spawn workers, STOP.
       Execute built-in Phase 1 (task discovery) -> role Phase 2-4 -> built-in Phase 5 (report).` }]
       })
       ```
-   d. Collect agent results: `wait_agent({ ids: [agentId], timeout_ms: 900000 })`
+   d. Collect agent results: `wait_agent({ targets: [taskId], timeout_ms: 900000 })`
    e. Read discoveries from output files
    f. Update tasks.json with results
-   g. Close agent: `close_agent({ id: agentId })`
+   g. Close agent: `close_agent({ target: taskId })`  // Use task_name, not agentId
 5. Parallel spawn rules:
 
 | Pipeline | Scenario | Spawn Behavior |
@@ -118,18 +129,48 @@ const agentIds = []
 for (const task of readyIdeatorTasks) {
   agentIds.push(spawn_agent({
     agent_type: "team_worker",
+    task_name: task.id,  // e.g., "IDEA-001" — enables named targeting
     items: [{ type: "text", text: `...role: ideator-<N>...` }]
   }))
 }
-const results = wait_agent({ ids: agentIds, timeout_ms: 900000 })
-// Process results, close agents
-for (const id of agentIds) { close_agent({ id }) }
+// Use task_name for stable targeting (v4)
+const taskNames = readyIdeatorTasks.map(t => t.id)
+const results = wait_agent({ targets: taskNames, timeout_ms: 900000 })
+if (results.timed_out) {
+  for (const taskId of taskNames) { state.tasks[taskId].status = 'timed_out'; close_agent({ target: taskId }) }
+} else {
+  // Process results, close agents
+  for (const taskId of taskNames) { close_agent({ target: taskId }) }
+}
 ```
 
+**Cross-Agent Supplementary Context** (v4):
+
+When spawning workers in a later pipeline phase, send upstream results as supplementary context to already-running workers:
+
+```
+// Example: Send ideation results to running challenger
+send_message({
+  target: "<running-agent-task-name>",
+  items: [{ type: "text", text: `## Supplementary Context\n${upstreamFindings}` }]
+})
+// Note: send_message queues info without interrupting the agent's current work
+```
+
+Use `send_message` (not `assign_task`) for supplementary info that enriches but doesn't redirect the agent's current task.
+
 6. Update session, output summary, STOP
 
 ## handleComplete
 
+**Cleanup Verification** (v4):
+```
+// Verify all agents are properly closed
+const remaining = list_agents({})
+// If any team agents still running -> close_agent each
+// Ensures clean session shutdown
+```
+
 Pipeline done. Generate report and completion action.
 
 Completion check by mode:

.codex/skills/team-brainstorm/roles/coordinator/role.md

@@ -6,7 +6,7 @@ Orchestrate team-brainstorm: topic clarify -> dispatch -> spawn -> monitor -> re
 
 **You are a dispatcher, not a doer.** Your ONLY outputs are:
 - Session state files (`.workflow/.team/` directory)
-- `spawn_agent` / `wait_agent` / `close_agent` / `send_input` calls
+- `spawn_agent` / `wait_agent` / `close_agent` / `send_message` / `assign_task` calls
 - Status reports to the user / `request_user_input` prompts
 
 **FORBIDDEN** (even if the task seems trivial):
@@ -34,6 +34,8 @@ WRONG: Edit/Write on project source files            — worker work
 - Manage Generator-Critic loop count (max 2 rounds)
 - Execute completion action in Phase 5
 - **Always proceed through full Phase 1-5 workflow, never skip to direct execution**
+- Use `send_message` for supplementary context (non-interrupting) and `assign_task` for triggering new work
+- Use `list_agents` for session resume health checks and cleanup verification
 
 ### MUST NOT
 - Generate ideas, challenge assumptions, synthesize, or evaluate -- workers handle this
@@ -149,6 +151,16 @@ Delegate to @commands/monitor.md#handleSpawnNext:
    - auto_archive -> Archive & Clean (status=completed, clean up session)
    - auto_keep -> Keep Active (status=paused)
 
+## v4 Coordination Patterns
+
+### Message Semantics
+- **send_message**: Queue supplementary info to a running agent. Does NOT interrupt current processing. Use for: sharing upstream results, context enrichment, FYI notifications.
+- **assign_task**: Assign new work and trigger processing. Use for: waking idle agents, redirecting work, requesting new output.
+
+### Agent Lifecycle Management
+- **list_agents({})**: Returns all running agents. Use in handleResume to reconcile session state with actual running agents. Use in handleComplete to verify clean shutdown.
+- **Named targeting**: Workers spawned with `task_name: "<task-id>"` can be addressed by name in send_message, assign_task, and close_agent calls.
+
 ## Error Handling
 
 | Error | Resolution |

.codex/skills/team-coordinate/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: team-coordinate
 description: Universal team coordination skill with dynamic role generation. Uses team-worker agent architecture with role-spec files. Only coordinator is built-in -- all worker roles are generated at runtime as role-specs and spawned via team-worker agent. Beat/cadence model for orchestration. Triggers on "Team Coordinate ".
-allowed-tools: spawn_agent(*), wait_agent(*), send_input(*), close_agent(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*)
+allowed-tools: spawn_agent(*), wait_agent(*), send_message(*), assign_task(*), close_agent(*), list_agents(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*)
 ---
 
 # Team Coordinate
@@ -40,7 +40,8 @@ Before calling ANY tool, apply this check:
 
 | Tool Call | Verdict | Reason |
 |-----------|---------|--------|
-| `spawn_agent`, `wait_agent`, `close_agent`, `send_input` | ALLOWED | Orchestration |
+| `spawn_agent`, `wait_agent`, `close_agent`, `send_message`, `assign_task` | ALLOWED | Orchestration |
+| `list_agents` | ALLOWED | Agent health check |
 | `request_user_input` | ALLOWED | User interaction |
 | `mcp__ccw-tools__team_msg` | ALLOWED | Message bus |
 | `Read/Write` on `.workflow/.team/` files | ALLOWED | Session state |
@@ -136,6 +137,8 @@ When coordinator spawns workers, use `team-worker` agent with role-spec path:
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
   items: [
     { type: "text", text: `## Role Assignment
 role: <role>
@@ -159,7 +162,7 @@ pipeline_phase: <pipeline-phase>` },
 })
 ```
 
-After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ id })` each worker.
+After spawning, use `wait_agent({ targets: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ target: <name> })` each worker.
 
 **Inner Loop roles** (role has 2+ serial same-prefix tasks): Set `inner_loop: true`. The team-worker agent handles the loop internally.
 
@@ -167,6 +170,71 @@ After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect
 
 ---
 
+
+### Model Selection Guide
+
+Roles are **dynamically generated** at runtime. Select model/reasoning_effort based on the generated role's `responsibility_type`:
+
+| responsibility_type | model | reasoning_effort | Rationale |
+|---------------------|-------|-------------------|-----------|
+| exploration | (default) | medium | Read-heavy, less reasoning needed |
+| analysis | (default) | high | Deep analysis requires full reasoning |
+| implementation | (default) | high | Code generation needs precision |
+| synthesis | (default) | medium | Aggregation over generation |
+| review | (default) | high | Quality assessment needs deep reasoning |
+
+Map each generated role's `responsibility_type` (from `team-session.json#roles`) to the table above.
+
+Override model/reasoning_effort in spawn_agent when cost optimization is needed:
+```
+spawn_agent({
+  agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
+  model: "<model-override>",
+  reasoning_effort: "<effort-level>",
+  items: [...]
+})
+```
+
+## v4 Agent Coordination
+
+### Message Semantics
+
+| Intent | API | Example |
+|--------|-----|---------|
+| Queue supplementary info (don't interrupt) | `send_message` | Send upstream task findings to a running downstream worker |
+| Not used in this skill | `assign_task` | No resident agents -- all workers are one-shot |
+| Check running agents | `list_agents` | Verify agent health during resume |
+
+**Note**: Since roles are dynamically generated, the coordinator must resolve task prefixes and role names from `team-session.json#roles` at runtime. There are no hardcoded role-specific examples.
+
+### fork_context Strategy
+
+`fork_context: false` is the default. Consider `fork_context: true` only when:
+- Runtime analysis reveals the task requires deep familiarity with the full conversation context
+- The dynamically-generated role-spec indicates the worker needs project-wide understanding
+- The coordinator has already accumulated significant context about the codebase
+
+This decision should be made per-task during Phase 4 based on the role's `responsibility_type`.
+
+### Agent Health Check
+
+Use `list_agents({})` in handleResume and handleComplete:
+
+```
+// Reconcile session state with actual running agents
+const running = list_agents({})
+// Compare with team-session.json active_workers
+// Reset orphaned tasks (in_progress but agent gone) to pending
+```
+
+### Named Agent Targeting
+
+Workers are spawned with `task_name: "<task-id>"` enabling direct addressing:
+- `send_message({ target: "<TASK-ID>", items: [...] })` -- queue upstream context without interrupting
+- `close_agent({ target: "<TASK-ID>" })` -- cleanup by name
+
 ## Completion Action
 
 When pipeline completes (all tasks done), coordinator presents an interactive choice:

.codex/skills/team-coordinate/roles/coordinator/commands/monitor.md

@@ -68,7 +68,7 @@ Receive result from wait_agent for [<role>]
   |   +- YES -> Update session state, do NOT remove from active_workers -> STOP
   +- Task status = completed?
   |   +- YES -> remove from active_workers -> update session
-  |   |   +- Close agent: close_agent({ id: <agentId> })
+  |   |   +- Close agent: close_agent({ target: <agentId> })
   |   |   +- -> handleSpawnNext
   |   +- NO -> progress message, do not advance -> STOP
   +- No matching worker found
@@ -114,6 +114,16 @@ Then STOP.
 
 ### Handler: handleResume
 
+**Agent Health Check** (v4):
+```
+// Verify actual running agents match session state
+const runningAgents = list_agents({})
+// For each active_agent in tasks.json:
+//   - If agent NOT in runningAgents -> agent crashed
+//   - Reset that task to pending, remove from active_agents
+// This prevents stale agent references from blocking the pipeline
+```
+
 Check active worker completion, process results, advance pipeline.
 
 ```
@@ -155,11 +165,27 @@ Ready tasks found?
       Update session file -> output summary -> STOP
 ```
 
+**Cross-Agent Supplementary Context** (v4):
+
+When spawning workers in a later pipeline phase, send upstream results as supplementary context to already-running workers:
+
+```
+// Example: Send upstream task results to running downstream workers
+send_message({
+  target: "<running-agent-task-name>",
+  items: [{ type: "text", text: `## Supplementary Context\n${upstreamFindings}` }]
+})
+// Note: send_message queues info without interrupting the agent's current work
+```
+
+Use `send_message` (not `assign_task`) for supplementary info that enriches but doesn't redirect the agent's current task.
+
 **Spawn worker call** (one per ready task):
 
 ```
 const agentId = spawn_agent({
   agent_type: "team_worker",
+  task_name: taskId,  // e.g., "EXPLORE-001" — enables named targeting
   items: [{ type: "text", text: `## Role Assignment
 role: <role>
 role_spec: <session-folder>/role-specs/<role>.md
@@ -171,16 +197,30 @@ inner_loop: <true|false>
 
 Read role_spec file to load Phase 2-4 domain instructions.` }]
 })
-// Collect results:
-const result = wait_agent({ ids: [agentId], timeout_ms: 900000 })
-// Process result, update tasks.json
-close_agent({ id: agentId })
+// Collect results — use task_name for stable targeting (v4):
+const result = wait_agent({ targets: [taskId], timeout_ms: 900000 })
+if (result.timed_out) {
+  state.tasks[taskId].status = 'timed_out'
+  close_agent({ target: taskId })
+  // Report timeout, STOP — let user decide retry
+} else {
+  // Process result, update tasks.json
+  close_agent({ target: taskId })  // Use task_name, not agentId
+}
 ```
 
 ---
 
 ### Handler: handleComplete
 
+**Cleanup Verification** (v4):
+```
+// Verify all agents are properly closed
+const remaining = list_agents({})
+// If any team agents still running -> close_agent each
+// Ensures clean session shutdown
+```
+
 Pipeline complete. Execute completion action based on session configuration.
 
 ```

.codex/skills/team-coordinate/roles/coordinator/role.md

@@ -31,7 +31,7 @@ OK: Write(".workflow/.team/TC-xxx/tasks.json", ...)   — task management
 OK: Read("roles/coordinator/commands/analyze-task.md") — own instructions
 OK: Read("specs/role-spec-template.md")               — generating role-specs
 OK: spawn_agent({ agent_type: "team_worker", ... })   — delegation
-OK: wait_agent({ ids: [...] })                        — monitoring
+OK: wait_agent({ targets: [...], timeout_ms: 900000 })     — monitoring
 ```
 
 **Self-check gate**: After Phase 1 analysis, before ANY other action, ask yourself:
@@ -57,6 +57,8 @@ OK: wait_agent({ ids: [...] })                        — monitoring
 - Handle consensus_blocked HIGH verdicts (create revision tasks or pause)
 - Detect fast-advance orphans on resume/check and reset to pending
 - Execute completion action when pipeline finishes
+- Use `send_message` for supplementary context (non-interrupting) and `assign_task` for triggering new work
+- Use `list_agents` for session resume health checks and cleanup verification
 
 ### MUST NOT
 - **Read source code or perform codebase exploration** (delegate to worker roles)
@@ -124,6 +126,7 @@ Phase 1 needs task analysis
 | tasks.json | File | Task lifecycle (create/read/update) |
 | team_msg | System | Message bus operations |
 | request_user_input | System | User interaction |
+| list_agents | System | Runtime agent discovery and health check |
 
 ---
 
@@ -373,6 +376,16 @@ Delegate to `@commands/dispatch.md` which creates the full task chain:
 
 ---
 
+## v4 Coordination Patterns
+
+### Message Semantics
+- **send_message**: Queue supplementary info to a running agent. Does NOT interrupt current processing. Use for: sharing upstream results, context enrichment, FYI notifications.
+- **assign_task**: Assign new work and trigger processing. Use for: waking idle agents, redirecting work, requesting new output.
+
+### Agent Lifecycle Management
+- **list_agents({})**: Returns all running agents. Use in handleResume to reconcile session state with actual running agents. Use in handleComplete to verify clean shutdown.
+- **Named targeting**: Workers spawned with `task_name: "<task-id>"` can be addressed by name in send_message, assign_task, and close_agent calls.
+
 ## Error Handling
 
 | Error | Resolution |

.codex/skills/team-designer/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: team-designer
 description: Meta-skill for generating team skills following the v4 architecture pattern. Produces complete skill packages with SKILL.md router, coordinator, worker roles, specs, and templates. Triggers on "team-designer", "design team".
-allowed-tools: spawn_agent(*), wait_agent(*), send_input(*), close_agent(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*)
+allowed-tools: spawn_agent(*), wait_agent(*), send_message(*), assign_task(*), close_agent(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*)
 ---
 
 # Team Skill Designer

.codex/skills/team-executor/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: team-executor
 description: Lightweight session execution skill. Resumes existing team-coordinate sessions for pure execution via team-worker agents. No analysis, no role generation -- only loads and executes. Session path required. Triggers on "Team Executor".
-allowed-tools: spawn_agent(*), wait_agent(*), send_input(*), close_agent(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*)
+allowed-tools: spawn_agent(*), wait_agent(*), send_message(*), assign_task(*), close_agent(*), list_agents(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*)
 ---
 
 # Team Executor
@@ -109,6 +109,8 @@ When executor spawns workers, use `team-worker` agent with role-spec path:
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
   items: [
     { type: "text", text: `## Role Assignment
 role: <role>
@@ -132,10 +134,35 @@ pipeline_phase: <pipeline-phase>` },
 })
 ```
 
-After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ id })` each worker.
+After spawning, use `wait_agent({ targets: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ target: <name> })` each worker.
 
 ---
 
+
+### Model Selection Guide
+
+team-executor loads roles dynamically from session role-specs. Use reasoning_effort based on the role type defined in the session:
+- Implementation/fix roles: `reasoning_effort: "high"`
+- Verification/test roles: `reasoning_effort: "medium"`
+- Default when role type is unclear: `reasoning_effort: "high"`
+
+## v4 Agent Coordination
+
+### State Reconciliation
+
+On resume, executor reconciles session state with actual running agents:
+```
+const running = list_agents({})
+// Compare with session's task-analysis.json active tasks
+// Reset orphaned tasks (in_progress but agent gone) to pending
+```
+
+### Worker Communication
+
+- `send_message({ target: "<task-id>", items: [...] })` -- queue supplementary context
+- `assign_task({ target: "<task-id>", items: [...] })` -- assign new work to inner_loop worker
+- `close_agent({ target: "<task-id>" })` -- cleanup completed worker
+
 ## Completion Action
 
 When pipeline completes (all tasks done), executor presents an interactive choice:

.codex/skills/team-executor/roles/executor/commands/monitor.md

@@ -167,22 +167,30 @@ After spawning all ready tasks:
 const agentIds = Object.values(state.active_agents)
   .filter(a => !a.resident)
   .map(a => a.agentId)
-wait_agent({ ids: agentIds, timeout_ms: 900000 })
-
-// Collect results from discoveries/{task_id}.json
-for (const [taskId, agent] of Object.entries(state.active_agents)) {
-  if (agent.resident) continue
-  try {
-    const disc = JSON.parse(Read(`${sessionFolder}/discoveries/${taskId}.json`))
-    state.tasks[taskId].status = disc.status || 'completed'
-    state.tasks[taskId].findings = disc.findings || ''
-    state.tasks[taskId].error = disc.error || null
-  } catch {
-    state.tasks[taskId].status = 'failed'
-    state.tasks[taskId].error = 'No discovery file produced'
+const waitResult = wait_agent({ targets: agentIds, timeout_ms: 900000 })
+if (waitResult.timed_out) {
+  for (const [taskId, agent] of Object.entries(state.active_agents)) {
+    if (agent.resident) continue
+    state.tasks[taskId].status = 'timed_out'
+    close_agent({ target: agent.agentId })
+    delete state.active_agents[taskId]
+  }
+} else {
+  // Collect results from discoveries/{task_id}.json
+  for (const [taskId, agent] of Object.entries(state.active_agents)) {
+    if (agent.resident) continue
+    try {
+      const disc = JSON.parse(Read(`${sessionFolder}/discoveries/${taskId}.json`))
+      state.tasks[taskId].status = disc.status || 'completed'
+      state.tasks[taskId].findings = disc.findings || ''
+      state.tasks[taskId].error = disc.error || null
+    } catch {
+      state.tasks[taskId].status = 'failed'
+      state.tasks[taskId].error = 'No discovery file produced'
+    }
+    close_agent({ target: agent.agentId })
+    delete state.active_agents[taskId]
   }
-  close_agent({ id: agent.agentId })
-  delete state.active_agents[taskId]
 }
 ```
 

.codex/skills/team-frontend-debug/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: team-frontend-debug
 description: Frontend debugging team using Chrome DevTools MCP. Dual-mode — feature-list testing or bug-report debugging. Triggers on "team-frontend-debug", "frontend debug".
-allowed-tools: spawn_agent(*), wait_agent(*), send_input(*), close_agent(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*), mcp__chrome-devtools__*(*)
+allowed-tools: spawn_agent(*), wait_agent(*), send_message(*), assign_task(*), close_agent(*), list_agents(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*), mcp__chrome-devtools__*(*)
 ---
 
 # Frontend Debug Team
@@ -64,7 +64,8 @@ Before calling ANY tool, apply this check:
 
 | Tool Call | Verdict | Reason |
 |-----------|---------|--------|
-| `spawn_agent`, `wait_agent`, `close_agent`, `send_input` | ALLOWED | Orchestration |
+| `spawn_agent`, `wait_agent`, `close_agent`, `send_message`, `assign_task` | ALLOWED | Orchestration |
+| `list_agents` | ALLOWED | Agent health check |
 | `request_user_input` | ALLOWED | User interaction |
 | `mcp__ccw-tools__team_msg` | ALLOWED | Message bus |
 | `Read/Write` on `.workflow/.team/` files | ALLOWED | Session state |
@@ -131,6 +132,8 @@ Coordinator spawns workers using this template:
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
   items: [
     { type: "text", text: `## Role Assignment
 role: <role>
@@ -154,7 +157,20 @@ pipeline_phase: <pipeline-phase>` },
 })
 ```
 
-After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ id })` each worker.
+After spawning, use `wait_agent({ targets: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ target })` each worker.
+
+
+### Model Selection Guide
+
+Debug workflows require tool-heavy interaction (Chrome DevTools MCP). Reasoning effort varies by role.
+
+| Role | reasoning_effort | Rationale |
+|------|-------------------|-----------|
+| tester | medium | Systematic feature testing via DevTools, follows test plan |
+| reproducer | medium | Reproduce bugs via DevTools interaction steps |
+| analyzer | high | Root cause analysis requires deep reasoning about evidence |
+| fixer | high | Code fixes must address root cause precisely |
+| verifier | medium | Verification follows defined success criteria via DevTools |
 
 ## User Commands
 
@@ -166,6 +182,53 @@ After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect
 | `feedback <text>` | Inject feedback for revision |
 | `retry <TASK-ID>` | Re-run a failed task |
 
+## v4 Agent Coordination
+
+### Message Semantics
+
+| Intent | API | Example |
+|--------|-----|---------|
+| Queue supplementary info (don't interrupt) | `send_message` | Send DevTools evidence to running analyzer |
+| Assign new work / trigger debug round | `assign_task` | Assign re-fix after verification failure |
+| Check running agents | `list_agents` | Verify agent health during resume |
+
+### Agent Health Check
+
+Use `list_agents({})` in handleResume and handleComplete:
+
+```
+// Reconcile session state with actual running agents
+const running = list_agents({})
+// Compare with team-session.json active tasks
+// Reset orphaned tasks (in_progress but agent gone) to pending
+```
+
+### Named Agent Targeting
+
+Workers are spawned with `task_name: "<task-id>"` enabling direct addressing:
+- `send_message({ target: "ANALYZE-001", items: [...] })` -- send evidence from reproducer to analyzer
+- `assign_task({ target: "FIX-001", items: [...] })` -- assign fix based on analysis results
+- `close_agent({ target: "VERIFY-001" })` -- cleanup after verification
+
+### Iterative Debug Loop Pattern
+
+When verifier reports a fix did not resolve the issue, coordinator uses `assign_task` to trigger re-analysis and re-fix:
+```
+// Verifier reports failure -> coordinator dispatches re-fix
+assign_task({
+  target: "FIX-001",   // reuse existing fixer if inner_loop, or spawn new
+  items: [
+    { type: "text", text: `## Re-fix Assignment
+verification_result: FAIL
+failure_evidence: <verifier's screenshot/console evidence>
+previous_fix_summary: <what was tried>
+instruction: Analyze verification failure and apply corrected fix.` }
+  ]
+})
+```
+
+This pattern enables iterative debug rounds: FIX -> VERIFY -> re-FIX -> re-VERIFY (max 3 rounds).
+
 ## Completion Action
 
 When pipeline completes, coordinator presents:

.codex/skills/team-frontend-debug/roles/coordinator/commands/monitor.md

@@ -67,6 +67,16 @@ Output:
 
 ## handleResume
 
+**Agent Health Check** (v4):
+```
+// Verify actual running agents match session state
+const runningAgents = list_agents({})
+// For each active_agent in tasks.json:
+//   - If agent NOT in runningAgents -> agent crashed
+//   - Reset that task to pending, remove from active_agents
+// This prevents stale agent references from blocking the pipeline
+```
+
 1. No active workers -> handleSpawnNext
 2. Has active -> check each status
    - completed -> mark done
@@ -90,6 +100,7 @@ Find ready tasks, spawn workers, STOP.
         ```
         spawn_agent({
           agent_type: "team_worker",
+          task_name: taskId,  // e.g., "TEST-001" — enables named targeting
           items: [{
             description: "Spawn <role> worker for <task-id>",
             team_name: "frontend-debug",
@@ -110,10 +121,33 @@ Find ready tasks, spawn workers, STOP.
         ```
       - Add to active_workers
 5. Update session, output summary, STOP
-6. Use `wait_agent({ ids: [<spawned-agent-ids>] })` to wait for callbacks. Workers use `report_agent_job_result()` to send results back.
+6. Use `wait_agent({ targets: [<spawned-task-names>], timeout_ms: 900000 })` to wait for callbacks. If `result.timed_out`, mark tasks as `timed_out` and close agents. Use `close_agent({ target: taskId })` with task_name for cleanup. Workers use `report_agent_job_result()` to send results back.
+
+**Cross-Agent Supplementary Context** (v4):
+
+When spawning workers in a later pipeline phase, send upstream results as supplementary context to already-running workers:
+
+```
+// Example: Send analysis results to running fixer
+send_message({
+  target: "<running-agent-task-name>",
+  items: [{ type: "text", text: `## Supplementary Context\n${upstreamFindings}` }]
+})
+// Note: send_message queues info without interrupting the agent's current work
+```
+
+Use `send_message` (not `assign_task`) for supplementary info that enriches but doesn't redirect the agent's current task.
 
 ## handleComplete
 
+**Cleanup Verification** (v4):
+```
+// Verify all agents are properly closed
+const remaining = list_agents({})
+// If any team agents still running -> close_agent each
+// Ensures clean session shutdown
+```
+
 Pipeline done. Generate debug report and completion action.
 
 1. Generate debug summary:

.codex/skills/team-frontend-debug/roles/coordinator/role.md

@@ -6,7 +6,7 @@ Orchestrate team-frontend-debug: analyze -> dispatch -> spawn -> monitor -> repo
 
 **You are a dispatcher, not a doer.** Your ONLY outputs are:
 - Session state files (`.workflow/.team/` directory)
-- `spawn_agent` / `wait_agent` / `close_agent` / `send_input` calls
+- `spawn_agent` / `wait_agent` / `close_agent` / `send_message` / `assign_task` calls
 - Status reports to the user / `request_user_input` prompts
 
 **FORBIDDEN** (even if the task seems trivial):
@@ -36,6 +36,8 @@ WRONG: mcp__chrome-devtools__* calls                 — worker work
 - Handle iteration loops (analyzer requesting more evidence)
 - Execute completion action when pipeline finishes
 - **Always proceed through full Phase 1-5 workflow, never skip to direct execution**
+- Use `send_message` for supplementary context (non-interrupting) and `assign_task` for triggering new work
+- Use `list_agents` for session resume health checks and cleanup verification
 
 ### MUST NOT
 - Read source code or explore codebase (delegate to workers)
@@ -141,6 +143,16 @@ Delegate to @commands/monitor.md#handleSpawnNext:
    - interactive -> request_user_input (Archive/Keep/Export)
    - auto_archive -> Archive & Clean
 
+## v4 Coordination Patterns
+
+### Message Semantics
+- **send_message**: Queue supplementary info to a running agent. Does NOT interrupt current processing. Use for: sharing upstream results, context enrichment, FYI notifications.
+- **assign_task**: Assign new work and trigger processing. Use for: waking idle agents, redirecting work, requesting new output.
+
+### Agent Lifecycle Management
+- **list_agents({})**: Returns all running agents. Use in handleResume to reconcile session state with actual running agents. Use in handleComplete to verify clean shutdown.
+- **Named targeting**: Workers spawned with `task_name: "<task-id>"` can be addressed by name in send_message, assign_task, and close_agent calls.
+
 ## Error Handling
 
 | Error | Resolution |

.codex/skills/team-frontend/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: team-frontend
 description: Unified team skill for frontend development. Pure router — all roles read this file. Beat model is coordinator-only in monitor.md. Built-in ui-ux-pro-max design intelligence. Triggers on "team frontend".
-allowed-tools: spawn_agent(*), wait_agent(*), send_input(*), close_agent(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*), WebFetch(*), WebSearch(*), mcp__ace-tool__search_context(*)
+allowed-tools: spawn_agent(*), wait_agent(*), send_message(*), assign_task(*), close_agent(*), list_agents(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*), WebFetch(*), WebSearch(*), mcp__ace-tool__search_context(*)
 ---
 
 # Team Frontend Development
@@ -54,7 +54,8 @@ Before calling ANY tool, apply this check:
 
 | Tool Call | Verdict | Reason |
 |-----------|---------|--------|
-| `spawn_agent`, `wait_agent`, `close_agent`, `send_input` | ALLOWED | Orchestration |
+| `spawn_agent`, `wait_agent`, `close_agent`, `send_message`, `assign_task` | ALLOWED | Orchestration |
+| `list_agents` | ALLOWED | Agent health check |
 | `request_user_input` | ALLOWED | User interaction |
 | `mcp__ccw-tools__team_msg` | ALLOWED | Message bus |
 | `Read/Write` on `.workflow/.team/` files | ALLOWED | Session state |
@@ -84,6 +85,8 @@ Coordinator spawns workers using this template:
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
   items: [
     { type: "text", text: `## Role Assignment
 role: <role>
@@ -107,7 +110,29 @@ pipeline_phase: <pipeline-phase>` },
 })
 ```
 
-After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ id })` each worker.
+After spawning, use `wait_agent({ targets: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ target })` each worker.
+
+
+### Model Selection Guide
+
+| Role | model | reasoning_effort | Rationale |
+|------|-------|-------------------|-----------|
+| Analyst (ANALYZE-*) | (default) | high | Design intelligence requires deep UI/UX reasoning |
+| Architect (ARCH-*) | (default) | high | Component spec and design token generation needs precision |
+| Developer (DEV-*) | (default) | high | Frontend code generation needs full reasoning |
+| QA (QA-*) | (default) | high | 5-dimension audit requires thorough analysis |
+
+Override model/reasoning_effort in spawn_agent when cost optimization is needed:
+```
+spawn_agent({
+  agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
+  model: "<model-override>",
+  reasoning_effort: "<effort-level>",
+  items: [...]
+})
+```
 
 ## User Commands
 
@@ -141,6 +166,55 @@ After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect
 
 - [specs/pipelines.md](specs/pipelines.md) — Pipeline definitions and task registry
 
+## v4 Agent Coordination
+
+### Message Semantics
+
+| Intent | API | Example |
+|--------|-----|---------|
+| Send architecture specs to running developer | `send_message` | Queue design tokens and component specs to DEV-* |
+| Not used in this skill | `assign_task` | No resident agents -- all workers are one-shot |
+| Check running agents | `list_agents` | Verify agent health during resume |
+
+### Pipeline Pattern
+
+Sequential pipeline with GC loops: analyst -> architect -> developer -> QA. In **system mode**, architect revision (ARCH-002) and developer (DEV-001) may run in parallel after QA-001 arch review.
+
+### fork_context Consideration
+
+Developer workers (DEV-*) benefit from `fork_context: true` when the coordinator has accumulated significant design context (analysis results, architecture specs, design tokens). This avoids the developer needing to re-read all upstream artifacts:
+
+```
+// Consider fork_context: true for developer in feature/system modes
+spawn_agent({
+  agent_type: "team_worker",
+  task_name: "DEV-001",
+  fork_context: true,  // Developer gets full coordinator context including design decisions
+  items: [...]
+})
+```
+
+### GC Loop (QA Fix Cycle)
+
+QA may flag FIX_REQUIRED, creating DEV-fix + QA-recheck tasks (max 2 rounds). If QA score remains < 6 after 2 rounds, escalate to user.
+
+### Agent Health Check
+
+Use `list_agents({})` in handleResume and handleComplete:
+
+```
+// Reconcile session state with actual running agents
+const running = list_agents({})
+// Compare with tasks.json active_agents
+// Reset orphaned tasks (in_progress but agent gone) to pending
+```
+
+### Named Agent Targeting
+
+Workers are spawned with `task_name: "<task-id>"` enabling direct addressing:
+- `send_message({ target: "DEV-001", items: [...] })` -- queue architecture specs to running developer
+- `close_agent({ target: "ARCH-001" })` -- cleanup by name after completion
+
 ## Error Handling
 
 | Scenario | Resolution |

.codex/skills/team-frontend/roles/coordinator/commands/monitor.md

@@ -109,6 +109,7 @@ Find and spawn the next ready tasks.
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: taskId,  // e.g., "DEV-001" — enables named targeting
   items: [{
     description: "Spawn <role> worker for <task-id>",
     team_name: "frontend",
@@ -136,7 +137,22 @@ Execute built-in Phase 1 -> role-spec Phase 2-4 -> built-in Phase 5.`
 | feature | After QA-001 arch review | Spawn DEV-001 |
 | system | After QA-001 arch review | Spawn ARCH-002 + DEV-001 in parallel |
 
-4. STOP after spawning -- use `wait_agent({ ids: [<spawned-agent-ids>] })` to wait for next callback
+4. STOP after spawning -- use `wait_agent({ targets: [<spawned-task-names>], timeout_ms: 900000 })` to wait for next callback. If `result.timed_out`, mark tasks as `timed_out` and close agents. Use `close_agent({ target: taskId })` with task_name for cleanup.
+
+**Cross-Agent Supplementary Context** (v4):
+
+When spawning workers in a later pipeline phase, send upstream results as supplementary context to already-running workers:
+
+```
+// Example: Send analysis results to running developer
+send_message({
+  target: "<running-agent-task-name>",
+  items: [{ type: "text", text: `## Supplementary Context\n${upstreamFindings}` }]
+})
+// Note: send_message queues info without interrupting the agent's current work
+```
+
+Use `send_message` (not `assign_task`) for supplementary info that enriches but doesn't redirect the agent's current task.
 
 ### handleCheck
 
@@ -157,6 +173,16 @@ Output status -- do NOT advance pipeline.
 
 ### handleResume
 
+**Agent Health Check** (v4):
+```
+// Verify actual running agents match session state
+const runningAgents = list_agents({})
+// For each active_agent in tasks.json:
+//   - If agent NOT in runningAgents -> agent crashed
+//   - Reset that task to pending, remove from active_agents
+// This prevents stale agent references from blocking the pipeline
+```
+
 Resume pipeline after user pause or interruption.
 
 1. Audit tasks.json for inconsistencies:
@@ -166,6 +192,14 @@ Resume pipeline after user pause or interruption.
 
 ### handleComplete
 
+**Cleanup Verification** (v4):
+```
+// Verify all agents are properly closed
+const remaining = list_agents({})
+// If any team agents still running -> close_agent each
+// Ensures clean session shutdown
+```
+
 Triggered when all pipeline tasks are completed.
 
 **Completion check**:

.codex/skills/team-frontend/roles/coordinator/role.md

@@ -6,7 +6,7 @@ Orchestrate team-frontend: analyze -> dispatch -> spawn -> monitor -> report.
 
 **You are a dispatcher, not a doer.** Your ONLY outputs are:
 - Session state files (`.workflow/.team/` directory)
-- `spawn_agent` / `wait_agent` / `close_agent` / `send_input` calls
+- `spawn_agent` / `wait_agent` / `close_agent` / `send_message` / `assign_task` calls
 - Status reports to the user / `request_user_input` prompts
 
 **FORBIDDEN** (even if the task seems trivial):
@@ -34,6 +34,8 @@ WRONG: Edit/Write on project source files            — worker work
 - Handle GC loops (developer <-> qa) with max 2 iterations
 - Execute completion action in Phase 5
 - **Always proceed through full Phase 1-5 workflow, never skip to direct execution**
+- Use `send_message` for supplementary context (non-interrupting) and `assign_task` for triggering new work
+- Use `list_agents` for session resume health checks and cleanup verification
 
 ### MUST NOT
 - Implement domain logic (analyzing, designing, coding, reviewing) -- workers handle this
@@ -165,6 +167,16 @@ Delegate to @commands/monitor.md#handleSpawnNext:
    - auto_archive -> Archive & Clean (status=completed, remove/archive session folder)
    - auto_keep -> Keep Active (status=paused)
 
+## v4 Coordination Patterns
+
+### Message Semantics
+- **send_message**: Queue supplementary info to a running agent. Does NOT interrupt current processing. Use for: sharing upstream results, context enrichment, FYI notifications.
+- **assign_task**: Assign new work and trigger processing. Use for: waking idle agents, redirecting work, requesting new output.
+
+### Agent Lifecycle Management
+- **list_agents({})**: Returns all running agents. Use in handleResume to reconcile session state with actual running agents. Use in handleComplete to verify clean shutdown.
+- **Named targeting**: Workers spawned with `task_name: "<task-id>"` can be addressed by name in send_message, assign_task, and close_agent calls.
+
 ## Error Handling
 
 | Error | Resolution |

.codex/skills/team-issue/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: team-issue
 description: Unified team skill for issue resolution. Uses team-worker agent architecture with role directories for domain logic. Coordinator orchestrates pipeline, workers are team-worker agents. Triggers on "team issue".
-allowed-tools: spawn_agent(*), wait_agent(*), send_input(*), close_agent(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*), mcp__ace-tool__search_context(*), mcp__ccw-tools__team_msg(*)
+allowed-tools: spawn_agent(*), wait_agent(*), send_message(*), assign_task(*), close_agent(*), list_agents(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*), mcp__ace-tool__search_context(*), mcp__ccw-tools__team_msg(*)
 ---
 
 # Team Issue Resolution
@@ -54,7 +54,8 @@ Before calling ANY tool, apply this check:
 
 | Tool Call | Verdict | Reason |
 |-----------|---------|--------|
-| `spawn_agent`, `wait_agent`, `close_agent`, `send_input` | ALLOWED | Orchestration |
+| `spawn_agent`, `wait_agent`, `close_agent`, `send_message`, `assign_task` | ALLOWED | Orchestration |
+| `list_agents` | ALLOWED | Agent health check |
 | `request_user_input` | ALLOWED | User interaction |
 | `mcp__ccw-tools__team_msg` | ALLOWED | Message bus |
 | `Read/Write` on `.workflow/.team/` files | ALLOWED | Session state |
@@ -85,6 +86,8 @@ Coordinator spawns workers using this template:
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
   items: [
     { type: "text", text: `## Role Assignment
 role: <role>
@@ -108,13 +111,15 @@ pipeline_phase: <pipeline-phase>` },
 })
 ```
 
-After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ id })` each worker.
+After spawning, use `wait_agent({ targets: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ target })` each worker.
 
 **Parallel spawn** (Batch mode, N explorer or M implementer instances):
 
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
   items: [
     { type: "text", text: `## Role Assignment
 role: <role>
@@ -139,7 +144,30 @@ pipeline_phase: <pipeline-phase>` },
 })
 ```
 
-After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ id })` each worker.
+After spawning, use `wait_agent({ targets: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ target })` each worker.
+
+
+### Model Selection Guide
+
+| Role | model | reasoning_effort | Rationale |
+|------|-------|-------------------|-----------|
+| Explorer (EXPLORE-*) | (default) | medium | Context gathering, file reading, less reasoning |
+| Planner (SOLVE-*) | (default) | high | Solution design requires deep analysis |
+| Reviewer (AUDIT-*) | (default) | high | Code review and plan validation need full reasoning |
+| Integrator (MARSHAL-*) | (default) | medium | Queue ordering and dependency resolution |
+| Implementer (BUILD-*) | (default) | high | Code generation needs precision |
+
+Override model/reasoning_effort in spawn_agent when cost optimization is needed:
+```
+spawn_agent({
+  agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
+  model: "<model-override>",
+  reasoning_effort: "<effort-level>",
+  items: [...]
+})
+```
 
 ## User Commands
 
@@ -176,6 +204,57 @@ After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect
 
 - [specs/pipelines.md](specs/pipelines.md) — Pipeline definitions and task registry
 
+## v4 Agent Coordination
+
+### Message Semantics
+
+| Intent | API | Example |
+|--------|-----|---------|
+| Send exploration context to running planner | `send_message` | Queue EXPLORE-* findings to SOLVE-* worker |
+| Not used in this skill | `assign_task` | No resident agents -- all workers are one-shot |
+| Check running agents | `list_agents` | Verify parallel explorer/implementer health |
+
+### Pipeline Pattern
+
+Pipeline with context passing: explore -> plan -> review (optional) -> marshal -> implement. In **Batch mode**, N explorers and M implementers run in parallel:
+
+```
+// Batch mode: spawn N explorers in parallel (max 5)
+const explorerNames = ["EXPLORE-001", "EXPLORE-002", ..., "EXPLORE-00N"]
+for (const name of explorerNames) {
+  spawn_agent({ agent_type: "team_worker", task_name: name, ... })
+}
+wait_agent({ targets: explorerNames, timeout_ms: 900000 })
+
+// After MARSHAL completes: spawn M implementers in parallel (max 3)
+const buildNames = ["BUILD-001", "BUILD-002", ..., "BUILD-00M"]
+for (const name of buildNames) {
+  spawn_agent({ agent_type: "team_worker", task_name: name, ... })
+}
+wait_agent({ targets: buildNames, timeout_ms: 900000 })
+```
+
+### Review-Fix Cycle
+
+Reviewer (AUDIT-*) may reject plans, triggering fix cycles (max 2). Dynamic SOLVE-fix and AUDIT re-review tasks are created in tasks.json.
+
+### Agent Health Check
+
+Use `list_agents({})` in handleResume and handleComplete:
+
+```
+// Reconcile session state with actual running agents
+const running = list_agents({})
+// Compare with tasks.json active_agents
+// Reset orphaned tasks (in_progress but agent gone) to pending
+```
+
+### Named Agent Targeting
+
+Workers are spawned with `task_name: "<task-id>"` enabling direct addressing:
+- `send_message({ target: "SOLVE-001", items: [...] })` -- queue exploration context to running planner
+- `close_agent({ target: "BUILD-001" })` -- cleanup by name after completion
+
 ## Error Handling
 
 | Scenario | Resolution |

.codex/skills/team-issue/roles/coordinator/commands/monitor.md

@@ -61,7 +61,7 @@ Worker completed. Process and advance.
    - Create BUILD-001..M tasks dynamically (add to tasks.json per dispatch.md Batch Pipeline BUILD section)
    - Proceed to handleSpawnNext
 
-6. Close completed agent: `close_agent({ id: <agentId> })`
+6. Close completed agent: `close_agent({ target: <agentId> })`
 7. Proceed to handleSpawnNext
 
 ## handleCheck
@@ -79,6 +79,16 @@ Read-only status report, then STOP.
 
 ## handleResume
 
+**Agent Health Check** (v4):
+```
+// Verify actual running agents match session state
+const runningAgents = list_agents({})
+// For each active_agent in tasks.json:
+//   - If agent NOT in runningAgents -> agent crashed
+//   - Reset that task to pending, remove from active_agents
+// This prevents stale agent references from blocking the pipeline
+```
+
 1. Audit task list: Tasks stuck in "in_progress" -> reset to "pending"
 2. Proceed to handleSpawnNext
 
@@ -96,6 +106,7 @@ Find ready tasks, spawn workers, STOP.
       ```
       const agentId = spawn_agent({
         agent_type: "team_worker",
+        task_name: taskId,  // e.g., "EXPLORE-001" — enables named targeting
         items: [{ type: "text", text: `## Role Assignment
       role: <role>
       role_spec: ~  or <project>/.codex/skills/team-issue/roles/<role>/role.md
@@ -109,10 +120,10 @@ Find ready tasks, spawn workers, STOP.
       Execute built-in Phase 1 (task discovery) -> role Phase 2-4 -> built-in Phase 5 (report).` }]
       })
       ```
-   d. Collect results: `wait_agent({ ids: [agentId], timeout_ms: 900000 })`
+   d. Collect results: `wait_agent({ targets: [taskId], timeout_ms: 900000 })`
    e. Read discoveries from output files
    f. Update tasks.json with results
-   g. Close agent: `close_agent({ id: agentId })`
+   g. Close agent: `close_agent({ target: taskId })`  // Use task_name, not agentId
 
 5. Parallel spawn rules:
 
@@ -130,6 +141,7 @@ const agentIds = []
 for (const task of readyTasks) {
   agentIds.push(spawn_agent({
     agent_type: "team_worker",
+    task_name: task.id,  // e.g., "EXPLORE-001" — enables named targeting
     items: [{ type: "text", text: `## Role Assignment
 role: <role>
 role_spec: ~  or <project>/.codex/skills/team-issue/roles/<role>/role.md
@@ -144,15 +156,44 @@ Read role_spec file to load Phase 2-4 domain instructions.
 Execute built-in Phase 1 (task discovery, owner=<role>-<N>) -> role Phase 2-4 -> built-in Phase 5 (report).` }]
   }))
 }
-const results = wait_agent({ ids: agentIds, timeout_ms: 900000 })
-// Process results, close agents
-for (const id of agentIds) { close_agent({ id }) }
+// Use task_name for stable targeting (v4)
+const taskNames = readyTasks.map(t => t.id)
+const results = wait_agent({ targets: taskNames, timeout_ms: 900000 })
+if (results.timed_out) {
+  for (const taskId of taskNames) { state.tasks[taskId].status = 'timed_out'; close_agent({ target: taskId }) }
+} else {
+  // Process results, close agents
+  for (const taskId of taskNames) { close_agent({ target: taskId }) }
+}
 ```
 
+**Cross-Agent Supplementary Context** (v4):
+
+When spawning workers in a later pipeline phase, send upstream results as supplementary context to already-running workers:
+
+```
+// Example: Send exploration results to running planner
+send_message({
+  target: "<running-agent-task-name>",
+  items: [{ type: "text", text: `## Supplementary Context\n${upstreamFindings}` }]
+})
+// Note: send_message queues info without interrupting the agent's current work
+```
+
+Use `send_message` (not `assign_task`) for supplementary info that enriches but doesn't redirect the agent's current task.
+
 6. Update session, output summary, STOP
 
 ## handleComplete
 
+**Cleanup Verification** (v4):
+```
+// Verify all agents are properly closed
+const remaining = list_agents({})
+// If any team agents still running -> close_agent each
+// Ensures clean session shutdown
+```
+
 Pipeline done. Generate report and completion action.
 
 Completion check by mode:

.codex/skills/team-issue/roles/coordinator/role.md

@@ -10,7 +10,7 @@ Orchestrate the issue resolution pipeline: clarify requirements -> create team -
 
 **You are a dispatcher, not a doer.** Your ONLY outputs are:
 - Session state files (`.workflow/.team/` directory)
-- `spawn_agent` / `wait_agent` / `close_agent` / `send_input` calls
+- `spawn_agent` / `wait_agent` / `close_agent` / `send_message` / `assign_task` calls
 - Status reports to the user / `request_user_input` prompts
 
 **FORBIDDEN** (even if the task seems trivial):
@@ -38,6 +38,8 @@ WRONG: Edit/Write on project source files            — worker work
 - Handle review-fix cycles with max 2 iterations
 - Execute completion action in Phase 5
 - **Always proceed through full Phase 1-5 workflow, never skip to direct execution**
+- Use `send_message` for supplementary context (non-interrupting) and `assign_task` for triggering new work
+- Use `list_agents` for session resume health checks and cleanup verification
 
 ### MUST NOT
 - Implement domain logic (exploring, planning, reviewing, implementing) -- workers handle this
@@ -183,6 +185,16 @@ Delegate to @commands/monitor.md#handleSpawnNext:
 | Keep Active | Update session status="paused" -> output: "Resume with: Skill(skill='team-issue', args='resume')" |
 | New Batch | Return to Phase 1 |
 
+## v4 Coordination Patterns
+
+### Message Semantics
+- **send_message**: Queue supplementary info to a running agent. Does NOT interrupt current processing. Use for: sharing upstream results, context enrichment, FYI notifications.
+- **assign_task**: Assign new work and trigger processing. Use for: waking idle agents, redirecting work, requesting new output.
+
+### Agent Lifecycle Management
+- **list_agents({})**: Returns all running agents. Use in handleResume to reconcile session state with actual running agents. Use in handleComplete to verify clean shutdown.
+- **Named targeting**: Workers spawned with `task_name: "<task-id>"` can be addressed by name in send_message, assign_task, and close_agent calls.
+
 ## Error Handling
 
 | Error | Resolution |

.codex/skills/team-iterdev/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: team-iterdev
 description: Unified team skill for iterative development team. Pure router — all roles read this file. Beat model is coordinator-only in monitor.md. Generator-Critic loops (developer<->reviewer, max 3 rounds). Triggers on "team iterdev".
-allowed-tools: spawn_agent(*), wait_agent(*), send_input(*), close_agent(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*)
+allowed-tools: spawn_agent(*), wait_agent(*), send_message(*), assign_task(*), close_agent(*), list_agents(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*)
 ---
 
 # Team IterDev
@@ -54,7 +54,8 @@ Before calling ANY tool, apply this check:
 
 | Tool Call | Verdict | Reason |
 |-----------|---------|--------|
-| `spawn_agent`, `wait_agent`, `close_agent`, `send_input` | ALLOWED | Orchestration |
+| `spawn_agent`, `wait_agent`, `close_agent`, `send_message`, `assign_task` | ALLOWED | Orchestration |
+| `list_agents` | ALLOWED | Agent health check |
 | `request_user_input` | ALLOWED | User interaction |
 | `mcp__ccw-tools__team_msg` | ALLOWED | Message bus |
 | `Read/Write` on `.workflow/.team/` files | ALLOWED | Session state |
@@ -84,6 +85,8 @@ Coordinator spawns workers using this template:
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
   items: [
     { type: "text", text: `## Role Assignment
 role: <role>
@@ -107,7 +110,19 @@ pipeline_phase: <pipeline-phase>` },
 })
 ```
 
-After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ id })` each worker.
+After spawning, use `wait_agent({ targets: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ target })` each worker.
+
+
+### Model Selection Guide
+
+Iterative development uses Generator-Critic loops. Developer and reviewer need high reasoning for code generation and review quality.
+
+| Role | reasoning_effort | Rationale |
+|------|-------------------|-----------|
+| architect | high | Design decisions require careful tradeoff analysis |
+| developer | high | Code generation needs precision, inner loop role |
+| tester | medium | Test execution follows defined verification plan |
+| reviewer | high | Critical review quality drives GC loop effectiveness |
 
 ## User Commands
 
@@ -145,6 +160,52 @@ After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect
 
 - [specs/pipelines.md](specs/pipelines.md) — Pipeline definitions and task registry
 
+## v4 Agent Coordination
+
+### Message Semantics
+
+| Intent | API | Example |
+|--------|-----|---------|
+| Queue supplementary info (don't interrupt) | `send_message` | Send design context to running developer |
+| Assign GC iteration round | `assign_task` | Assign revision to developer after reviewer feedback |
+| Check running agents | `list_agents` | Verify agent health during resume |
+
+### Agent Health Check
+
+Use `list_agents({})` in handleResume and handleComplete:
+
+```
+// Reconcile session state with actual running agents
+const running = list_agents({})
+// Compare with task-ledger.json active tasks
+// Reset orphaned tasks (in_progress but agent gone) to pending
+```
+
+### Named Agent Targeting
+
+Workers are spawned with `task_name: "<task-id>"` enabling direct addressing:
+- `send_message({ target: "DEV-001", items: [...] })` -- send design context to developer
+- `assign_task({ target: "DEV-001", items: [...] })` -- assign revision after reviewer feedback
+- `close_agent({ target: "REVIEW-001" })` -- cleanup after GC loop completes
+
+### Generator-Critic Loop via assign_task (Deep Interaction Pattern)
+
+The developer-reviewer GC loop uses `assign_task` for iteration rounds:
+```
+// Round N: Reviewer found issues -> coordinator assigns revision to developer
+assign_task({
+  target: "DEV-001",
+  items: [
+    { type: "text", text: `## GC Revision Round ${N}
+review_feedback: <reviewer findings from REVIEW-001>
+iteration: ${N}/3
+instruction: Address reviewer feedback. Focus on: <specific issues>.` }
+  ]
+})
+```
+
+After developer completes revision, coordinator spawns/assigns reviewer for next round. Max 3 rounds per GC cycle.
+
 ## Error Handling
 
 | Scenario | Resolution |

.codex/skills/team-iterdev/roles/coordinator/commands/monitor.md

@@ -93,6 +93,7 @@ state.tasks[taskId].status = 'in_progress'
 // 2) Spawn worker
 const agentId = spawn_agent({
   agent_type: "team_worker",
+  task_name: taskId,  // e.g., "DEV-001" — enables named targeting
   items: [
     { type: "text", text: `## Role Assignment
 role: ${role}
@@ -110,12 +111,19 @@ Execute built-in Phase 1 -> role-spec Phase 2-4 -> built-in Phase 5.` }
 // 3) Track agent
 state.active_agents[taskId] = { agentId, role, started_at: now }
 
-// 4) Wait for completion
-wait_agent({ ids: [agentId] })
-
-// 5) Collect results and update tasks.json
-state.tasks[taskId].status = 'completed'
-delete state.active_agents[taskId]
+// 4) Wait for completion — use task_name for stable targeting (v4)
+const waitResult = wait_agent({ targets: [taskId], timeout_ms: 900000 })
+if (waitResult.timed_out) {
+  state.tasks[taskId].status = 'timed_out'
+  close_agent({ target: taskId })
+  delete state.active_agents[taskId]
+  // Report timeout, STOP
+} else {
+  // 5) Collect results and update tasks.json
+  state.tasks[taskId].status = 'completed'
+  close_agent({ target: taskId })  // Use task_name, not agentId
+  delete state.active_agents[taskId]
+}
 ```
 
 4. **Parallel spawn rules**:
@@ -128,6 +136,21 @@ delete state.active_agents[taskId]
 | Multi-Sprint | VERIFY + DEV-fix both unblocked | Spawn BOTH in parallel, wait_agent for both |
 | Multi-Sprint | Other stages | One worker at a time |
 
+**Cross-Agent Supplementary Context** (v4):
+
+When spawning workers in a later pipeline phase, send upstream results as supplementary context to already-running workers:
+
+```
+// Example: Send design results to running developer
+send_message({
+  target: "<running-agent-task-name>",
+  items: [{ type: "text", text: `## Supplementary Context\n${upstreamFindings}` }]
+})
+// Note: send_message queues info without interrupting the agent's current work
+```
+
+Use `send_message` (not `assign_task`) for supplementary info that enriches but doesn't redirect the agent's current task.
+
 5. STOP after processing -- wait for next event
 
 ### handleCheck
@@ -149,6 +172,16 @@ Session: <session-id>
 
 ### handleResume
 
+**Agent Health Check** (v4):
+```
+// Verify actual running agents match session state
+const runningAgents = list_agents({})
+// For each active_agent in tasks.json:
+//   - If agent NOT in runningAgents -> agent crashed
+//   - Reset that task to pending, remove from active_agents
+// This prevents stale agent references from blocking the pipeline
+```
+
 Resume pipeline after user pause or interruption.
 
 1. Audit tasks.json for inconsistencies:
@@ -158,6 +191,14 @@ Resume pipeline after user pause or interruption.
 
 ### handleComplete
 
+**Cleanup Verification** (v4):
+```
+// Verify all agents are properly closed
+const remaining = list_agents({})
+// If any team agents still running -> close_agent each
+// Ensures clean session shutdown
+```
+
 Triggered when all pipeline tasks are completed.
 
 **Completion check by mode**:

.codex/skills/team-iterdev/roles/coordinator/role.md

@@ -6,7 +6,7 @@ Orchestrate team-iterdev: analyze -> dispatch -> spawn -> monitor -> report.
 
 **You are a dispatcher, not a doer.** Your ONLY outputs are:
 - Session state files (`.workflow/.team/` directory)
-- `spawn_agent` / `wait_agent` / `close_agent` / `send_input` calls
+- `spawn_agent` / `wait_agent` / `close_agent` / `send_message` / `assign_task` calls
 - Status reports to the user / `request_user_input` prompts
 
 **FORBIDDEN** (even if the task seems trivial):
@@ -35,6 +35,8 @@ WRONG: Edit/Write on project source files            — worker work
 - Maintain tasks.json for real-time progress
 - Execute completion action in Phase 5
 - **Always proceed through full Phase 1-5 workflow, never skip to direct execution**
+- Use `send_message` for supplementary context (non-interrupting) and `assign_task` for triggering new work
+- Use `list_agents` for session resume health checks and cleanup verification
 
 ### MUST NOT
 - Implement domain logic (designing, coding, testing, reviewing) -- workers handle this
@@ -169,6 +171,16 @@ Delegate to @commands/monitor.md#handleSpawnNext:
    - auto_archive -> Archive & Clean (status=completed)
    - auto_keep -> Keep Active (status=paused)
 
+## v4 Coordination Patterns
+
+### Message Semantics
+- **send_message**: Queue supplementary info to a running agent. Does NOT interrupt current processing. Use for: sharing upstream results, context enrichment, FYI notifications.
+- **assign_task**: Assign new work and trigger processing. Use for: waking idle agents, redirecting work, requesting new output.
+
+### Agent Lifecycle Management
+- **list_agents({})**: Returns all running agents. Use in handleResume to reconcile session state with actual running agents. Use in handleComplete to verify clean shutdown.
+- **Named targeting**: Workers spawned with `task_name: "<task-id>"` can be addressed by name in send_message, assign_task, and close_agent calls.
+
 ## Error Handling
 
 | Error | Resolution |

.codex/skills/team-lifecycle-v4/MIGRATION-PLAN.md

@@ -19,7 +19,7 @@
 | Worker spawn | `Agent({ subagent_type: "team-worker", prompt })` | `spawn_agent({ agent_type: "tlv4_worker", items })` |
 | Supervisor spawn | `Agent({ subagent_type: "team-supervisor", prompt })` | `spawn_agent({ agent_type: "tlv4_supervisor", items })` |
 | Supervisor wake | `SendMessage({ recipient: "supervisor", content })` | `send_input({ id: supervisorId, items })` |
-| Supervisor shutdown | `SendMessage({ type: "shutdown_request" })` | `close_agent({ id: supervisorId })` |
+| Supervisor shutdown | `SendMessage({ type: "shutdown_request" })` | `close_agent({ target: supervisorId })` |
 | 等待完成 | 后台回调 -> monitor.md | `wait_agent({ ids, timeout_ms })` |
 | 任务状态 | `TaskCreate` / `TaskUpdate` | `tasks.json` 文件读写 |
 | 团队管理 | `TeamCreate` / `TeamDelete` | session folder init / cleanup |
@@ -222,14 +222,14 @@ scope: [${task.deps}]
 pipeline_progress: ${done}/${total} tasks completed` }
   ]
 })
-wait_agent({ ids: [supervisorId], timeout_ms: 300000 })
+wait_agent({ targets: [supervisorId], timeout_ms: 300000 })
 ```
 
 ### Supervisor Shutdown
 
 ```javascript
 // 对齐 Claude Code SendMessage({ type: "shutdown_request" })
-close_agent({ id: supervisorId })
+close_agent({ target: supervisorId })
 ```
 
 ### Wave 执行引擎
@@ -300,7 +300,7 @@ pipeline_phase: ${task.pipeline_phase}` },
 
   // 2) 批量等待
   if (agentMap.length > 0) {
-    wait_agent({ ids: agentMap.map(a => a.agentId), timeout_ms: 900000 })
+    wait_agent({ targets: agentMap.map(a => a.agentId), timeout_ms: 900000 })
   }
 
   // 3) 收集结果，合并到 tasks.json
@@ -315,7 +315,7 @@ pipeline_phase: ${task.pipeline_phase}` },
       state.tasks[taskId].status = 'failed'
       state.tasks[taskId].error = 'No discovery file produced'
     }
-    close_agent({ id: agentId })
+    close_agent({ target: agentId })
   }
 
   // 4) 执行 CHECKPOINT 任务 (send_input 唤醒 supervisor)
@@ -329,7 +329,7 @@ scope: [${task.deps.join(', ')}]
 pipeline_progress: ${completedCount}/${totalCount} tasks completed` }
       ]
     })
-    wait_agent({ ids: [supervisorId], timeout_ms: 300000 })
+    wait_agent({ targets: [supervisorId], timeout_ms: 300000 })
 
     // 读取 checkpoint 报告
     try {

.codex/skills/team-lifecycle-v4/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: team-lifecycle-v4
 description: Full lifecycle team skill with clean architecture. SKILL.md is a universal router — all roles read it. Beat model is coordinator-only. Structure is roles/ + specs/ + templates/. Triggers on "team lifecycle v4".
-allowed-tools: spawn_agent(*), wait_agent(*), send_input(*), close_agent(*), report_agent_job_result(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*), request_user_input(*)
+allowed-tools: spawn_agent(*), wait_agent(*), send_message(*), assign_task(*), close_agent(*), list_agents(*), report_agent_job_result(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*), request_user_input(*)
 ---
 
 # Team Lifecycle v4
@@ -29,7 +29,7 @@ Skill(skill="team-lifecycle-v4", args="task description")
             spawn_agent    ...     spawn_agent
           (team_worker)         (team_supervisor)
               per-task             resident agent
-              lifecycle            send_input-driven
+              lifecycle            assign_task-driven
                     |                     |
                     +-- wait_agent --------+
                               |
@@ -63,7 +63,8 @@ Before calling ANY tool, apply this check:
 
 | Tool Call | Verdict | Reason |
 |-----------|---------|--------|
-| `spawn_agent`, `wait_agent`, `close_agent`, `send_input` | ALLOWED | Orchestration |
+| `spawn_agent`, `wait_agent`, `close_agent`, `send_message`, `assign_task` | ALLOWED | Orchestration |
+| `list_agents` | ALLOWED | Agent health check |
 | `request_user_input` | ALLOWED | User interaction |
 | `mcp__ccw-tools__team_msg` | ALLOWED | Message bus |
 | `Read/Write` on `.workflow/.team/` files | ALLOWED | Session state |
@@ -94,6 +95,8 @@ Coordinator spawns workers using this template:
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
   items: [
     { type: "text", text: `## Role Assignment
 role: <role>
@@ -120,13 +123,15 @@ pipeline_phase: <pipeline-phase>` },
 
 ## Supervisor Spawn Template
 
-Supervisor is a **resident agent** (independent from team_worker). Spawned once during session init, woken via send_input for each CHECKPOINT task.
+Supervisor is a **resident agent** (independent from team_worker). Spawned once during session init, woken via assign_task for each CHECKPOINT task.
 
 ### Spawn (Phase 2 -- once per session)
 
 ```
 supervisorId = spawn_agent({
   agent_type: "team_supervisor",
+  task_name: "supervisor",
+  fork_context: false,
   items: [
     { type: "text", text: `## Role Assignment
 role: supervisor
@@ -137,7 +142,7 @@ requirement: <task-description>
 
 Read role_spec file (<skill_root>/roles/supervisor/role.md) to load checkpoint definitions.
 Init: load baseline context, report ready, go idle.
-Wake cycle: orchestrator sends checkpoint requests via send_input.` }
+Wake cycle: orchestrator sends checkpoint requests via assign_task.` }
   ]
 })
 ```
@@ -145,8 +150,8 @@ Wake cycle: orchestrator sends checkpoint requests via send_input.` }
 ### Wake (per CHECKPOINT task)
 
 ```
-send_input({
-  id: supervisorId,
+assign_task({
+  target: "supervisor",
   items: [
     { type: "text", text: `## Checkpoint Request
 task_id: <CHECKPOINT-NNN>
@@ -154,13 +159,38 @@ scope: [<upstream-task-ids>]
 pipeline_progress: <done>/<total> tasks completed` }
   ]
 })
-wait_agent({ ids: [supervisorId], timeout_ms: 300000 })
+wait_agent({ targets: ["supervisor"], timeout_ms: 300000 })
 ```
 
 ### Shutdown (pipeline complete)
 
 ```
-close_agent({ id: supervisorId })
+close_agent({ target: "supervisor" })
+```
+
+
+### Model Selection Guide
+
+| Role | model | reasoning_effort | Rationale |
+|------|-------|-------------------|-----------|
+| Analyst (RESEARCH-*) | (default) | medium | Read-heavy exploration, less reasoning needed |
+| Writer (DRAFT-*) | (default) | high | Spec writing requires precision and completeness |
+| Planner (PLAN-*) | (default) | high | Architecture decisions need full reasoning |
+| Executor (IMPL-*) | (default) | high | Code generation needs precision |
+| Tester (TEST-*) | (default) | high | Test generation requires deep code understanding |
+| Reviewer (REVIEW-*, QUALITY-*, IMPROVE-*) | (default) | high | Deep analysis for quality assessment |
+| Supervisor (CHECKPOINT-*) | (default) | medium | Gate checking, report aggregation |
+
+Override model/reasoning_effort in spawn_agent when cost optimization is needed:
+```
+spawn_agent({
+  agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
+  model: "<model-override>",
+  reasoning_effort: "<effort-level>",
+  items: [...]
+})
 ```
 
 ## Wave Execution Engine
@@ -172,9 +202,9 @@ For each wave in the pipeline:
 3. **Build upstream context** -- For each task, gather findings from `context_from` tasks via tasks.json and `discoveries/{id}.json`
 4. **Separate task types** -- Split into regular tasks and CHECKPOINT tasks
 5. **Spawn regular tasks** -- For each regular task, call `spawn_agent({ agent_type: "team_worker", items: [...] })`, collect agent IDs
-6. **Wait** -- `wait_agent({ ids: [...], timeout_ms: 900000 })`
-7. **Collect results** -- Read `discoveries/{task_id}.json` for each agent, update tasks.json status/findings/error, then `close_agent({ id })` each worker
-8. **Execute checkpoints** -- For each CHECKPOINT task, `send_input` to supervisor, `wait_agent`, read checkpoint report from `artifacts/`, parse verdict
+6. **Wait** -- `wait_agent({ targets: [...], timeout_ms: 900000 })`
+7. **Collect results** -- Read `discoveries/{task_id}.json` for each agent, update tasks.json status/findings/error, then `close_agent({ target })` each worker
+8. **Execute checkpoints** -- For each CHECKPOINT task, `assign_task` to supervisor, `wait_agent`, read checkpoint report from `artifacts/`, parse verdict
 9. **Handle block** -- If verdict is `block`, prompt user via `request_user_input` with options: Override / Revise upstream / Abort
 10. **Persist** -- Write updated state to `<session>/tasks.json`
 
@@ -189,6 +219,39 @@ For each wave in the pipeline:
 | `recheck` | Re-run quality check |
 | `improve [dimension]` | Auto-improve weakest dimension |
 
+## v4 Agent Coordination
+
+### Message Semantics
+
+| Intent | API | Example |
+|--------|-----|---------|
+| Queue supplementary info (don't interrupt) | `send_message` | Send planning results to running implementers |
+| Wake resident supervisor for checkpoint | `assign_task` | Trigger CHECKPOINT-* evaluation on supervisor |
+| Supervisor reports back to coordinator | `send_message` | Supervisor sends checkpoint verdict as supplementary info |
+| Check running agents | `list_agents` | Verify agent + supervisor health during resume |
+
+**CRITICAL**: The supervisor is a **resident agent** woken via `assign_task`, NOT `send_message`. Regular workers complete and are closed; the supervisor persists across checkpoints. See "Supervisor Spawn Template" above.
+
+### Agent Health Check
+
+Use `list_agents({})` in handleResume and handleComplete:
+
+```
+// Reconcile session state with actual running agents
+const running = list_agents({})
+// Compare with tasks.json active_agents
+// Reset orphaned tasks (in_progress but agent gone) to pending
+// ALSO check supervisor: if supervisor missing but CHECKPOINT tasks pending -> respawn
+```
+
+### Named Agent Targeting
+
+Workers are spawned with `task_name: "<task-id>"` enabling direct addressing:
+- `send_message({ target: "IMPL-001", items: [...] })` -- queue planning context to running implementer
+- `assign_task({ target: "supervisor", items: [...] })` -- wake supervisor for checkpoint
+- `close_agent({ target: "IMPL-001" })` -- cleanup regular worker by name
+- `close_agent({ target: "supervisor" })` -- shutdown supervisor at pipeline end
+
 ## Completion Action
 
 When pipeline completes, coordinator presents:

.codex/skills/team-lifecycle-v4/roles/coordinator/commands/dispatch.md

@@ -48,7 +48,7 @@ CHECKPOINT tasks are dispatched like regular tasks but handled differently at sp
 
 - Added to tasks.json with proper deps (upstream tasks that must complete first)
 - Owner: supervisor
-- **NOT spawned as tlv4_worker** — coordinator wakes the resident supervisor via send_input
+- **NOT spawned as tlv4_worker** — coordinator wakes the resident supervisor via assign_task
 - If `supervision: false` in tasks.json, skip creating CHECKPOINT tasks entirely
 - RoleSpec in description: `<project>/.codex/skills/team-lifecycle-v4/roles/supervisor/role.md`
 

.codex/skills/team-lifecycle-v4/roles/coordinator/commands/monitor.md

@@ -5,7 +5,7 @@ Synchronous pipeline coordination using spawn_agent + wait_agent.
 ## Constants
 
 - WORKER_AGENT: tlv4_worker
-- SUPERVISOR_AGENT: tlv4_supervisor (resident, woken via send_input)
+- SUPERVISOR_AGENT: tlv4_supervisor (resident, woken via assign_task)
 
 ## Handler Router
 
@@ -35,6 +35,16 @@ Output:
 
 ## handleResume
 
+**Agent Health Check** (v4):
+```
+// Verify actual running agents match session state
+const runningAgents = list_agents({})
+// For each active_agent in tasks.json:
+//   - If agent NOT in runningAgents -> agent crashed
+//   - Reset that task to pending, remove from active_agents
+// This prevents stale agent references from blocking the pipeline
+```
+
 1. Read tasks.json, check active_agents
 2. No active agents -> handleSpawnNext
 3. Has active agents -> check each:
@@ -63,6 +73,7 @@ state.tasks[task.id].status = 'in_progress'
 // 2) Spawn worker
 const agentId = spawn_agent({
   agent_type: "tlv4_worker",
+  task_name: task.id,  // e.g., "PLAN-001" — enables named targeting
   items: [
     { type: "text", text: `## Role Assignment
 role: ${task.role}
@@ -92,30 +103,52 @@ state.active_agents[task.id] = { agentId, role: task.role, started_at: now }
 After spawning all ready regular tasks:
 
 ```javascript
-// 4) Batch wait for all spawned workers
-const agentIds = Object.values(state.active_agents)
-  .filter(a => !a.resident)
-  .map(a => a.agentId)
-wait_agent({ ids: agentIds, timeout_ms: 900000 })
-
-// 5) Collect results from discoveries/{task_id}.json
-for (const [taskId, agent] of Object.entries(state.active_agents)) {
-  if (agent.resident) continue
-  try {
-    const disc = JSON.parse(Read(`${sessionFolder}/discoveries/${taskId}.json`))
-    state.tasks[taskId].status = disc.status || 'completed'
-    state.tasks[taskId].findings = disc.findings || ''
-    state.tasks[taskId].quality_score = disc.quality_score || null
-    state.tasks[taskId].error = disc.error || null
-  } catch {
-    state.tasks[taskId].status = 'failed'
-    state.tasks[taskId].error = 'No discovery file produced'
+// 4) Batch wait — use task_name for stable targeting (v4)
+const taskNames = Object.entries(state.active_agents)
+  .filter(([_, a]) => !a.resident)
+  .map(([taskId]) => taskId)
+const waitResult = wait_agent({ targets: taskNames, timeout_ms: 900000 })
+if (waitResult.timed_out) {
+  for (const taskId of taskNames) {
+    state.tasks[taskId].status = 'timed_out'
+    close_agent({ target: taskId })
+    delete state.active_agents[taskId]
+  }
+} else {
+  // 5) Collect results from discoveries/{task_id}.json
+  for (const [taskId, agent] of Object.entries(state.active_agents)) {
+    if (agent.resident) continue
+    try {
+      const disc = JSON.parse(Read(`${sessionFolder}/discoveries/${taskId}.json`))
+      state.tasks[taskId].status = disc.status || 'completed'
+      state.tasks[taskId].findings = disc.findings || ''
+      state.tasks[taskId].quality_score = disc.quality_score || null
+      state.tasks[taskId].error = disc.error || null
+    } catch {
+      state.tasks[taskId].status = 'failed'
+      state.tasks[taskId].error = 'No discovery file produced'
+    }
+    close_agent({ target: taskId })  // Use task_name, not agentId
+    delete state.active_agents[taskId]
   }
-  close_agent({ id: agent.agentId })
-  delete state.active_agents[taskId]
 }
 ```
 
+**Cross-Agent Supplementary Context** (v4):
+
+When spawning workers in a later pipeline phase, send upstream results as supplementary context to already-running workers:
+
+```
+// Example: Send planning results to running implementers
+send_message({
+  target: "<running-agent-task-name>",
+  items: [{ type: "text", text: `## Supplementary Context\n${upstreamFindings}` }]
+})
+// Note: send_message queues info without interrupting the agent's current work
+```
+
+Use `send_message` (not `assign_task`) for supplementary info that enriches but doesn't redirect the agent's current task.
+
 ### Handle CHECKPOINT Tasks
 
 For each ready CHECKPOINT task:
@@ -125,7 +158,7 @@ For each ready CHECKPOINT task:
 2. Determine scope: list task IDs that this checkpoint depends on (its deps)
 3. Wake supervisor:
    ```javascript
-   send_input({
+   assign_task({
      id: supervisorId,
      items: [
        { type: "text", text: `## Checkpoint Request
@@ -134,7 +167,8 @@ For each ready CHECKPOINT task:
    pipeline_progress: ${completedCount}/${totalCount} tasks completed` }
      ]
    })
-   wait_agent({ ids: [supervisorId], timeout_ms: 300000 })
+   const cpResult = wait_agent({ targets: [supervisorId], timeout_ms: 300000 })
+   if (cpResult.timed_out) { /* mark checkpoint timed_out, close supervisor, STOP */ }
    ```
 4. Read checkpoint report from artifacts/${task.id}-report.md
 5. Parse verdict (pass / warn / block):
@@ -153,11 +187,19 @@ After processing all results:
 
 ## handleComplete
 
+**Cleanup Verification** (v4):
+```
+// Verify all agents are properly closed
+const remaining = list_agents({})
+// If any team agents still running -> close_agent each
+// Ensures clean session shutdown
+```
+
 Pipeline done. Generate report and completion action.
 
 1. Shutdown resident supervisor (if active):
    ```javascript
-   close_agent({ id: supervisorId })
+   close_agent({ target: supervisorId })
    ```
    Remove from active_agents in tasks.json
 2. Generate summary (deliverables, stats, discussions)

.codex/skills/team-lifecycle-v4/roles/coordinator/role.md

@@ -6,7 +6,7 @@ Orchestrate team-lifecycle-v4: analyze -> dispatch -> spawn -> monitor -> report
 
 **You are a dispatcher, not a doer.** Your ONLY outputs are:
 - Session state files (`.workflow/.team/` directory)
-- `spawn_agent` / `wait_agent` / `close_agent` / `send_input` calls
+- `spawn_agent` / `wait_agent` / `close_agent` / `send_message` / `assign_task` calls
 - Status reports to the user
 - `request_user_input` prompts
 
@@ -38,6 +38,8 @@ WRONG: Bash("npm test"), Bash("tsc"), etc.           — worker work
 - Maintain session state (tasks.json)
 - Handle capability_gap reports
 - Execute completion action when pipeline finishes
+- Use `send_message` for supplementary context (non-interrupting) and `assign_task` for triggering new work
+- Use `list_agents` for session resume health checks and cleanup verification
 
 ### MUST NOT
 - Read source code or explore codebase (delegate to workers)
@@ -166,6 +168,16 @@ Delegate to @commands/monitor.md#handleSpawnNext:
    - auto_archive -> Archive & Clean (rm -rf session folder)
    - auto_keep -> Keep Active
 
+## v4 Coordination Patterns
+
+### Message Semantics
+- **send_message**: Queue supplementary info to a running agent. Does NOT interrupt current processing. Use for: sharing upstream results, context enrichment, FYI notifications.
+- **assign_task**: Assign new work and trigger processing. Use for: waking idle agents, redirecting work, requesting new output.
+
+### Agent Lifecycle Management
+- **list_agents({})**: Returns all running agents. Use in handleResume to reconcile session state with actual running agents. Use in handleComplete to verify clean shutdown.
+- **Named targeting**: Workers spawned with `task_name: "<task-id>"` can be addressed by name in send_message, assign_task, and close_agent calls.
+
 ## Error Handling
 
 | Error | Resolution |

.codex/skills/team-lifecycle-v4/roles/supervisor/role.md

@@ -17,7 +17,7 @@ Process and execution supervision at pipeline phase transition points.
 ## Identity
 - Tag: [supervisor] | Prefix: CHECKPOINT-*
 - Responsibility: Verify cross-artifact consistency, process compliance, and execution health between pipeline phases
-- Residency: Spawned once, awakened via `send_input` at each checkpoint trigger (not SendMessage)
+- Residency: Spawned once, awakened via `assign_task` at each checkpoint trigger (not SendMessage)
 
 ## Boundaries
 

.codex/skills/team-perf-opt/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: team-perf-opt
 description: Unified team skill for performance optimization. Coordinator orchestrates pipeline, workers are team-worker agents. Supports single/fan-out/independent parallel modes. Triggers on "team perf-opt".
-allowed-tools: spawn_agent(*), wait_agent(*), send_input(*), close_agent(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*), mcp__ace-tool__search_context(*)
+allowed-tools: spawn_agent(*), wait_agent(*), send_message(*), assign_task(*), close_agent(*), list_agents(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*), mcp__ace-tool__search_context(*)
 ---
 
 # Team Performance Optimization
@@ -65,7 +65,8 @@ Before calling ANY tool, apply this check:
 
 | Tool Call | Verdict | Reason |
 |-----------|---------|--------|
-| `spawn_agent`, `wait_agent`, `close_agent`, `send_input` | ALLOWED | Orchestration |
+| `spawn_agent`, `wait_agent`, `close_agent`, `send_message`, `assign_task` | ALLOWED | Orchestration |
+| `list_agents` | ALLOWED | Agent health check |
 | `request_user_input` | ALLOWED | User interaction |
 | `mcp__ccw-tools__team_msg` | ALLOWED | Message bus |
 | `Read/Write` on `.workflow/.team/` files | ALLOWED | Session state |
@@ -96,6 +97,8 @@ Coordinator spawns workers using this template:
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
   items: [
     { type: "text", text: `## Role Assignment
 role: <role>
@@ -119,11 +122,37 @@ pipeline_phase: <pipeline-phase>` },
 })
 ```
 
-After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ id })` each worker.
+After spawning, use `wait_agent({ targets: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ target })` each worker.
 
 **Inner Loop roles** (optimizer): Set `inner_loop: true`.
 **Single-task roles** (profiler, strategist, benchmarker, reviewer): Set `inner_loop: false`.
 
+
+### Model Selection Guide
+
+Performance optimization is measurement-driven. Profiler and benchmarker need consistent context for before/after comparison.
+
+| Role | reasoning_effort | Rationale |
+|------|-------------------|-----------|
+| profiler | high | Must identify subtle bottlenecks from profiling data |
+| strategist | high | Optimization strategy requires understanding tradeoffs |
+| optimizer | high | Performance-critical code changes need precision |
+| benchmarker | medium | Benchmark execution follows defined measurement plan |
+| reviewer | high | Must verify optimizations don't introduce regressions |
+
+### Benchmark Context Sharing with fork_context
+
+For before/after comparison, benchmarker should share context with profiler's baseline:
+```
+spawn_agent({
+  agent_type: "team_worker",
+  task_name: "BENCH-001",
+  fork_context: true,   // Share context so benchmarker sees profiler's baseline metrics
+  reasoning_effort: "medium",
+  items: [...]
+})
+```
+
 ## User Commands
 
 | Command | Action |
@@ -155,6 +184,42 @@ After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect
 +-- .msg/meta.json                  # Session metadata
 ```
 
+## v4 Agent Coordination
+
+### Message Semantics
+
+| Intent | API | Example |
+|--------|-----|---------|
+| Queue supplementary info (don't interrupt) | `send_message` | Send baseline metrics to running optimizer |
+| Assign fix after benchmark regression | `assign_task` | Assign FIX task when benchmark shows regression |
+| Check running agents | `list_agents` | Verify agent health during resume |
+
+### Agent Health Check
+
+Use `list_agents({})` in handleResume and handleComplete:
+
+```
+// Reconcile session state with actual running agents
+const running = list_agents({})
+// Compare with session.json active tasks
+// Reset orphaned tasks (in_progress but agent gone) to pending
+```
+
+### Named Agent Targeting
+
+Workers are spawned with `task_name: "<task-id>"` enabling direct addressing:
+- `send_message({ target: "IMPL-001", items: [...] })` -- send strategy details to optimizer
+- `assign_task({ target: "IMPL-001", items: [...] })` -- assign fix after benchmark regression
+- `close_agent({ target: "BENCH-001" })` -- cleanup after benchmarking completes
+
+### Baseline-to-Result Pipeline
+
+Profiler baseline metrics flow through the pipeline and must reach benchmarker for comparison:
+1. PROFILE-001 produces `baseline-metrics.json` in artifacts/
+2. Coordinator includes baseline reference in upstream context for all downstream workers
+3. BENCH-001 reads baseline and compares against post-optimization measurements
+4. If regression detected, coordinator auto-creates FIX task with regression details
+
 ## Completion Action
 
 When the pipeline completes:

.codex/skills/team-perf-opt/roles/coordinator/commands/monitor.md

@@ -65,6 +65,7 @@ Find and spawn the next ready tasks.
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: taskId,  // e.g., "IMPL-001" — enables named targeting
   items: [{
     description: "Spawn <role> worker for <task-id>",
     team_name: "perf-opt",
@@ -93,7 +94,22 @@ Execute built-in Phase 1 -> role-spec Phase 2-4 -> built-in Phase 5.`
 | Fan-out (IMPL-B{NN} done) | BENCH-B{NN} + REVIEW-B{NN} ready | Spawn both for that branch in parallel |
 | Independent | Any unblocked task | Spawn all ready tasks across all pipelines in parallel |
 
-4. STOP after spawning -- use `wait_agent({ ids: [<spawned-agent-ids>] })` to wait for next callback
+4. STOP after spawning -- use `wait_agent({ targets: [<spawned-task-names>], timeout_ms: 900000 })` to wait for next callback. If `result.timed_out`, mark tasks as `timed_out` and close agents. Use `close_agent({ target: taskId })` with task_name for cleanup.
+
+**Cross-Agent Supplementary Context** (v4):
+
+When spawning workers in a later pipeline phase, send upstream results as supplementary context to already-running workers:
+
+```
+// Example: Send profiling results to running optimizers
+send_message({
+  target: "<running-agent-task-name>",
+  items: [{ type: "text", text: `## Supplementary Context\n${upstreamFindings}` }]
+})
+// Note: send_message queues info without interrupting the agent's current work
+```
+
+Use `send_message` (not `assign_task`) for supplementary info that enriches but doesn't redirect the agent's current task.
 
 ### Review-Fix Cycle (CP-3)
 
@@ -242,6 +258,16 @@ Output status -- do NOT advance pipeline.
 
 ### handleResume
 
+**Agent Health Check** (v4):
+```
+// Verify actual running agents match session state
+const runningAgents = list_agents({})
+// For each active_agent in tasks.json:
+//   - If agent NOT in runningAgents -> agent crashed
+//   - Reset that task to pending, remove from active_agents
+// This prevents stale agent references from blocking the pipeline
+```
+
 Resume pipeline after user pause or interruption.
 
 1. Audit tasks.json for inconsistencies:
@@ -262,6 +288,14 @@ Handle consensus_blocked signals from discuss rounds.
 
 ### handleComplete
 
+**Cleanup Verification** (v4):
+```
+// Verify all agents are properly closed
+const remaining = list_agents({})
+// If any team agents still running -> close_agent each
+// Ensures clean session shutdown
+```
+
 Triggered when all pipeline tasks are completed and no fix cycles remain.
 
 **Completion check varies by mode**:

.codex/skills/team-perf-opt/roles/coordinator/role.md

@@ -10,7 +10,7 @@ Orchestrates the performance optimization pipeline: manages task chains, spawns
 
 **You are a dispatcher, not a doer.** Your ONLY outputs are:
 - Session state files (`.workflow/.team/` directory)
-- `spawn_agent` / `wait_agent` / `close_agent` / `send_input` calls
+- `spawn_agent` / `wait_agent` / `close_agent` / `send_message` / `assign_task` calls
 - Status reports to the user / `request_user_input` prompts
 
 **FORBIDDEN** (even if the task seems trivial):
@@ -35,6 +35,8 @@ WRONG: Edit/Write on project source files            — worker work
 - Handle review-fix cycles with max 3 iterations per branch
 - Execute completion action in Phase 5
 - **Always proceed through full Phase 1-5 workflow, never skip to direct execution**
+- Use `send_message` for supplementary context (non-interrupting) and `assign_task` for triggering new work
+- Use `list_agents` for session resume health checks and cleanup verification
 
 ### MUST NOT
 
@@ -155,6 +157,16 @@ All subsequent coordination handled by `@commands/monitor.md`.
 
 ---
 
+## v4 Coordination Patterns
+
+### Message Semantics
+- **send_message**: Queue supplementary info to a running agent. Does NOT interrupt current processing. Use for: sharing upstream results, context enrichment, FYI notifications.
+- **assign_task**: Assign new work and trigger processing. Use for: waking idle agents, redirecting work, requesting new output.
+
+### Agent Lifecycle Management
+- **list_agents({})**: Returns all running agents. Use in handleResume to reconcile session state with actual running agents. Use in handleComplete to verify clean shutdown.
+- **Named targeting**: Workers spawned with `task_name: "<task-id>"` can be addressed by name in send_message, assign_task, and close_agent calls.
+
 ## Error Handling
 
 | Scenario | Resolution |

.codex/skills/team-planex/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: team-planex
 description: Unified team skill for plan-and-execute pipeline. Pure router — coordinator always. Beat model is coordinator-only in monitor.md. Triggers on "team planex".
-allowed-tools: spawn_agent(*), wait_agent(*), send_input(*), close_agent(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*), mcp__ccw-tools__team_msg(*)
+allowed-tools: spawn_agent(*), wait_agent(*), send_message(*), assign_task(*), close_agent(*), list_agents(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*), mcp__ccw-tools__team_msg(*)
 ---
 
 # Team PlanEx
@@ -53,7 +53,8 @@ Before calling ANY tool, apply this check:
 
 | Tool Call | Verdict | Reason |
 |-----------|---------|--------|
-| `spawn_agent`, `wait_agent`, `close_agent`, `send_input` | ALLOWED | Orchestration |
+| `spawn_agent`, `wait_agent`, `close_agent`, `send_message`, `assign_task` | ALLOWED | Orchestration |
+| `list_agents` | ALLOWED | Agent health check |
 | `request_user_input` | ALLOWED | User interaction |
 | `mcp__ccw-tools__team_msg` | ALLOWED | Message bus |
 | `Read/Write` on `.workflow/.team/` files | ALLOWED | Session state |
@@ -83,6 +84,8 @@ Coordinator spawns workers using this template:
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
   items: [
     { type: "text", text: `## Role Assignment
 role: <role>
@@ -107,7 +110,27 @@ pipeline_phase: <pipeline-phase>` },
 })
 ```
 
-After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ id })` each worker.
+After spawning, use `wait_agent({ targets: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ target })` each worker.
+
+
+### Model Selection Guide
+
+| Role | model | reasoning_effort | Rationale |
+|------|-------|-------------------|-----------|
+| Planner (PLAN-*) | (default) | high | Solution planning requires deep code analysis |
+| Executor (EXEC-*) | (default) | high | Code implementation needs precision |
+
+Override model/reasoning_effort in spawn_agent when cost optimization is needed:
+```
+spawn_agent({
+  agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
+  model: "<model-override>",
+  reasoning_effort: "<effort-level>",
+  items: [...]
+})
+```
 
 ## User Commands
 
@@ -140,6 +163,45 @@ After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect
 
 - [specs/pipelines.md](specs/pipelines.md) — Pipeline definitions, task metadata registry, execution method selection
 
+## v4 Agent Coordination
+
+### Message Semantics
+
+| Intent | API | Example |
+|--------|-----|---------|
+| Send plan updates to running executor | `send_message` | Queue planner solution details to EXEC-* workers |
+| Not used in this skill | `assign_task` | No resident agents -- planner and executor are one-shot |
+| Check running agents | `list_agents` | Verify planner/executor health during resume |
+
+### Two-Phase Pipeline Pattern
+
+Plan-and-execute is a **Two-Phase pattern**: planner creates solution plans (PLAN-*), then coordinator spawns executors (EXEC-*) for each planned issue. The planner may dynamically create EXEC-* task entries in tasks.json.
+
+```
+// Phase 1: Planner runs, creates EXEC-* tasks in tasks.json
+spawn_agent({ agent_type: "team_worker", task_name: "PLAN-001", ... })
+wait_agent({ targets: ["PLAN-001"], timeout_ms: 900000 })
+// Phase 2: Executors run per-issue, may run in sequence or parallel
+// Inner loop: planner/executor handle multiple tasks internally
+```
+
+### Agent Health Check
+
+Use `list_agents({})` in handleResume and handleComplete:
+
+```
+// Reconcile session state with actual running agents
+const running = list_agents({})
+// Compare with tasks.json active_agents
+// Reset orphaned tasks (in_progress but agent gone) to pending
+```
+
+### Named Agent Targeting
+
+Workers are spawned with `task_name: "<task-id>"` enabling direct addressing:
+- `send_message({ target: "EXEC-001", items: [...] })` -- queue plan solution to running executor
+- `close_agent({ target: "PLAN-001" })` -- cleanup by name after completion
+
 ## Error Handling
 
 | Scenario | Resolution |

.codex/skills/team-planex/roles/coordinator/commands/monitor.md

@@ -42,7 +42,7 @@ Receive result from wait_agent for [<role>]
   |   +- YES -> Update session -> STOP
   +- Task status = completed?
   |   +- YES -> remove from active_workers -> update session
-  |   |   +- Close agent: close_agent({ id: <agentId> })
+  |   |   +- Close agent: close_agent({ target: <agentId> })
   |   |   +- role = planner?
   |   |   |   +- Check for new EXEC-* tasks in tasks.json (planner creates them)
   |   |   |   +- -> handleSpawnNext (spawn executor for new EXEC-* tasks)
@@ -87,6 +87,16 @@ Then STOP.
 
 ### Handler: handleResume
 
+**Agent Health Check** (v4):
+```
+// Verify actual running agents match session state
+const runningAgents = list_agents({})
+// For each active_agent in tasks.json:
+//   - If agent NOT in runningAgents -> agent crashed
+//   - Reset that task to pending, remove from active_agents
+// This prevents stale agent references from blocking the pipeline
+```
+
 ```
 Load active_workers
   +- No active workers -> handleSpawnNext
@@ -120,6 +130,7 @@ Collect task states from tasks.json
       +- Spawn team_worker:
          const agentId = spawn_agent({
            agent_type: "team_worker",
+           task_name: taskId,  // e.g., "PLAN-001" — enables named targeting
            items: [{ type: "text", text: `## Role Assignment
 role: <role>
 role_spec: ~  or <project>/.codex/skills/team-planex/roles/<role>/role.md
@@ -130,14 +141,49 @@ requirement: <task-description>
 inner_loop: true
 execution_method: <method>` }]
          })
-         // Collect results:
-         const result = wait_agent({ ids: [agentId], timeout_ms: 900000 })
-         // Process result, update tasks.json
-         close_agent({ id: agentId })
+         // Collect results — use task_name for stable targeting (v4):
+         const result = wait_agent({ targets: [taskId], timeout_ms: 900000 })
+         if (result.timed_out) {
+           state.tasks[taskId].status = 'timed_out'
+           close_agent({ target: taskId })
+           // Report timeout, STOP
+         } else {
+           // Process result, update tasks.json
+           close_agent({ target: taskId })  // Use task_name, not agentId
+         }
       +- Add to session.active_workers
       Update session -> output summary -> STOP
 ```
 
+**Cross-Agent Supplementary Context** (v4):
+
+When spawning workers in a later pipeline phase, send upstream results as supplementary context to already-running workers:
+
+```
+// Example: Send planning results to running executors
+send_message({
+  target: "<running-agent-task-name>",
+  items: [{ type: "text", text: `## Supplementary Context\n${upstreamFindings}` }]
+})
+// Note: send_message queues info without interrupting the agent's current work
+```
+
+Use `send_message` (not `assign_task`) for supplementary info that enriches but doesn't redirect the agent's current task.
+
+---
+
+### Pipeline Complete (PIPELINE_COMPLETE -> Phase 5)
+
+**Cleanup Verification** (v4):
+```
+// Verify all agents are properly closed
+const remaining = list_agents({})
+// If any team agents still running -> close_agent each
+// Ensures clean session shutdown
+```
+
+When all tasks are completed (no pending, no in_progress), transition to coordinator Phase 5.
+
 ---
 
 ## Phase 4: Validation

.codex/skills/team-planex/roles/coordinator/role.md

@@ -6,7 +6,7 @@ Orchestrate team-planex: analyze -> dispatch -> spawn -> monitor -> report.
 
 **You are a dispatcher, not a doer.** Your ONLY outputs are:
 - Session state files (`.workflow/.team/` directory)
-- `spawn_agent` / `wait_agent` / `close_agent` / `send_input` calls
+- `spawn_agent` / `wait_agent` / `close_agent` / `send_message` / `assign_task` calls
 - Status reports to the user / `request_user_input` prompts
 
 **FORBIDDEN** (even if the task seems trivial):
@@ -33,6 +33,8 @@ WRONG: Edit/Write on project source files            — worker work
 - Monitor progress via `commands/monitor.md` with Spawn-and-Stop pattern
 - Maintain session state (.msg/meta.json)
 - **Always proceed through full Phase 1-5 workflow, never skip to direct execution**
+- Use `send_message` for supplementary context (non-interrupting) and `assign_task` for triggering new work
+- Use `list_agents` for session resume health checks and cleanup verification
 
 ### MUST NOT
 - Execute planning or implementation work directly (delegate to workers)
@@ -148,6 +150,16 @@ Delegate to `@commands/dispatch.md`:
    - auto_keep -> Keep Active (status=paused)
    - auto_yes -> Archive & Clean without prompting
 
+## v4 Coordination Patterns
+
+### Message Semantics
+- **send_message**: Queue supplementary info to a running agent. Does NOT interrupt current processing. Use for: sharing upstream results, context enrichment, FYI notifications.
+- **assign_task**: Assign new work and trigger processing. Use for: waking idle agents, redirecting work, requesting new output.
+
+### Agent Lifecycle Management
+- **list_agents({})**: Returns all running agents. Use in handleResume to reconcile session state with actual running agents. Use in handleComplete to verify clean shutdown.
+- **Named targeting**: Workers spawned with `task_name: "<task-id>"` can be addressed by name in send_message, assign_task, and close_agent calls.
+
 ## Error Handling
 
 | Error | Resolution |

.codex/skills/team-quality-assurance/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: team-quality-assurance
 description: Unified team skill for quality assurance. Full closed-loop QA combining issue discovery and software testing. Triggers on "team quality-assurance", "team qa".
-allowed-tools: spawn_agent(*), wait_agent(*), send_input(*), close_agent(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*)
+allowed-tools: spawn_agent(*), wait_agent(*), send_message(*), assign_task(*), close_agent(*), list_agents(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*)
 ---
 
 # Team Quality Assurance
@@ -55,7 +55,8 @@ Before calling ANY tool, apply this check:
 
 | Tool Call | Verdict | Reason |
 |-----------|---------|--------|
-| `spawn_agent`, `wait_agent`, `close_agent`, `send_input` | ALLOWED | Orchestration |
+| `spawn_agent`, `wait_agent`, `close_agent`, `send_message`, `assign_task` | ALLOWED | Orchestration |
+| `list_agents` | ALLOWED | Agent health check |
 | `request_user_input` | ALLOWED | User interaction |
 | `mcp__ccw-tools__team_msg` | ALLOWED | Message bus |
 | `Read/Write` on `.workflow/.team/` files | ALLOWED | Session state |
@@ -86,6 +87,8 @@ Coordinator spawns workers using this template:
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
   items: [
     { type: "text", text: `## Role Assignment
 role: <role>
@@ -109,7 +112,30 @@ pipeline_phase: <pipeline-phase>` },
 })
 ```
 
-After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ id })` each worker.
+After spawning, use `wait_agent({ targets: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ target })` each worker.
+
+
+### Model Selection Guide
+
+| Role | model | reasoning_effort | Rationale |
+|------|-------|-------------------|-----------|
+| Scout (SCOUT-*) | (default) | medium | Issue discovery scanning, less reasoning needed |
+| Strategist (QASTRAT-*) | (default) | high | Test strategy design requires deep analysis |
+| Generator (QAGEN-*) | (default) | high | Test code generation needs precision |
+| Executor (QARUN-*) | (default) | medium | Running tests and collecting results |
+| Analyst (QAANA-*) | (default) | high | Quality analysis and coverage assessment |
+
+Override model/reasoning_effort in spawn_agent when cost optimization is needed:
+```
+spawn_agent({
+  agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
+  model: "<model-override>",
+  reasoning_effort: "<effort-level>",
+  items: [...]
+})
+```
 
 ## User Commands
 
@@ -121,6 +147,37 @@ After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect
 | `--mode=testing` | Force testing mode |
 | `--mode=full` | Force full QA mode |
 
+## v4 Agent Coordination
+
+### Message Semantics
+
+| Intent | API | Example |
+|--------|-----|---------|
+| Send scout findings to running strategist | `send_message` | Queue issue scan results to QASTRAT-* |
+| Not used in this skill | `assign_task` | No resident agents -- all workers are one-shot |
+| Check running agents | `list_agents` | Verify agent health during resume |
+
+### Pipeline Pattern
+
+Sequential pipeline with GC loops: scout -> strategist -> generator -> executor -> analyst. The executor/generator may loop via GC fix tasks when coverage is below target (max 3 rounds).
+
+### Agent Health Check
+
+Use `list_agents({})` in handleResume and handleComplete:
+
+```
+// Reconcile session state with actual running agents
+const running = list_agents({})
+// Compare with tasks.json active_agents
+// Reset orphaned tasks (in_progress but agent gone) to pending
+```
+
+### Named Agent Targeting
+
+Workers are spawned with `task_name: "<task-id>"` enabling direct addressing:
+- `send_message({ target: "QASTRAT-001", items: [...] })` -- queue scout findings to running strategist
+- `close_agent({ target: "SCOUT-001" })` -- cleanup by name after completion
+
 ## Completion Action
 
 When pipeline completes, coordinator presents:

.codex/skills/team-quality-assurance/roles/coordinator/commands/monitor.md

@@ -95,6 +95,16 @@ Then STOP.
 
 ## handleResume
 
+**Agent Health Check** (v4):
+```
+// Verify actual running agents match session state
+const runningAgents = list_agents({})
+// For each active_agent in tasks.json:
+//   - If agent NOT in runningAgents -> agent crashed
+//   - Reset that task to pending, remove from active_agents
+// This prevents stale agent references from blocking the pipeline
+```
+
 1. No active workers -> handleSpawnNext
 2. Has active -> check each status
    - completed -> mark done (update tasks.json)
@@ -132,6 +142,7 @@ Find ready tasks, spawn workers, STOP.
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: taskId,  // e.g., "SCOUT-001" — enables named targeting
   items: [{
     description: "Spawn <role> worker for <subject>",
     team_name: "quality-assurance",
@@ -157,10 +168,33 @@ Execute built-in Phase 1 (task discovery) -> role Phase 2-4 -> built-in Phase 5
 
    f. Add to active_workers
 5. Update session, output summary, STOP
-6. Use `wait_agent({ ids: [<spawned-agent-ids>] })` to wait for callbacks. Workers use `report_agent_job_result()` to send results back.
+6. Use `wait_agent({ targets: [<spawned-task-names>], timeout_ms: 900000 })` to wait for callbacks. If `result.timed_out`, mark tasks as `timed_out` and close agents. Use `close_agent({ target: taskId })` with task_name for cleanup. Workers use `report_agent_job_result()` to send results back.
+
+**Cross-Agent Supplementary Context** (v4):
+
+When spawning workers in a later pipeline phase, send upstream results as supplementary context to already-running workers:
+
+```
+// Example: Send scout results to running strategist
+send_message({
+  target: "<running-agent-task-name>",
+  items: [{ type: "text", text: `## Supplementary Context\n${upstreamFindings}` }]
+})
+// Note: send_message queues info without interrupting the agent's current work
+```
+
+Use `send_message` (not `assign_task`) for supplementary info that enriches but doesn't redirect the agent's current task.
 
 ## handleComplete
 
+**Cleanup Verification** (v4):
+```
+// Verify all agents are properly closed
+const remaining = list_agents({})
+// If any team agents still running -> close_agent each
+// Ensures clean session shutdown
+```
+
 Pipeline done. Generate report and completion action.
 
 1. Verify all tasks (including GC fix/recheck tasks) have status "completed" or "deleted" in tasks.json

.codex/skills/team-quality-assurance/roles/coordinator/role.md

@@ -6,7 +6,7 @@ Orchestrate team-quality-assurance: analyze -> dispatch -> spawn -> monitor -> r
 
 **You are a dispatcher, not a doer.** Your ONLY outputs are:
 - Session state files (`.workflow/.team/` directory)
-- `spawn_agent` / `wait_agent` / `close_agent` / `send_input` calls
+- `spawn_agent` / `wait_agent` / `close_agent` / `send_message` / `assign_task` calls
 - Status reports to the user / `request_user_input` prompts
 
 **FORBIDDEN** (even if the task seems trivial):
@@ -36,6 +36,8 @@ WRONG: Bash("npm test"), Bash("tsc"), etc.           — worker work
 - Handle GC loop (generator-executor coverage cycles)
 - Execute completion action when pipeline finishes
 - **Always proceed through full Phase 1-5 workflow, never skip to direct execution**
+- Use `send_message` for supplementary context (non-interrupting) and `assign_task` for triggering new work
+- Use `list_agents` for session resume health checks and cleanup verification
 
 ### MUST NOT
 - Read source code or explore codebase (delegate to workers)
@@ -151,6 +153,16 @@ Delegate to @commands/monitor.md#handleSpawnNext:
    - auto_archive -> Archive & Clean
    - auto_keep -> Keep Active
 
+## v4 Coordination Patterns
+
+### Message Semantics
+- **send_message**: Queue supplementary info to a running agent. Does NOT interrupt current processing. Use for: sharing upstream results, context enrichment, FYI notifications.
+- **assign_task**: Assign new work and trigger processing. Use for: waking idle agents, redirecting work, requesting new output.
+
+### Agent Lifecycle Management
+- **list_agents({})**: Returns all running agents. Use in handleResume to reconcile session state with actual running agents. Use in handleComplete to verify clean shutdown.
+- **Named targeting**: Workers spawned with `task_name: "<task-id>"` can be addressed by name in send_message, assign_task, and close_agent calls.
+
 ## Error Handling
 
 | Error | Resolution |

.codex/skills/team-review/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: team-review
 description: "Unified team skill for code review. 3-role pipeline: scanner, reviewer, fixer. Triggers on team-review."
-allowed-tools: spawn_agent(*), wait_agent(*), send_input(*), close_agent(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*), mcp__ace-tool__search_context(*)
+allowed-tools: spawn_agent(*), wait_agent(*), send_message(*), assign_task(*), close_agent(*), list_agents(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*), mcp__ace-tool__search_context(*)
 ---
 
 # Team Review
@@ -53,7 +53,8 @@ Before calling ANY tool, apply this check:
 
 | Tool Call | Verdict | Reason |
 |-----------|---------|--------|
-| `spawn_agent`, `wait_agent`, `close_agent`, `send_input` | ALLOWED | Orchestration |
+| `spawn_agent`, `wait_agent`, `close_agent`, `send_message`, `assign_task` | ALLOWED | Orchestration |
+| `list_agents` | ALLOWED | Agent health check |
 | `request_user_input` | ALLOWED | User interaction |
 | `mcp__ccw-tools__team_msg` | ALLOWED | Message bus |
 | `Read/Write` on `.workflow/.team/` files | ALLOWED | Session state |
@@ -84,6 +85,8 @@ Coordinator spawns workers using this template:
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
   items: [
     { type: "text", text: `## Role Assignment
 role: <role>
@@ -107,7 +110,28 @@ pipeline_phase: <pipeline-phase>` },
 })
 ```
 
-After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ id })` each worker.
+After spawning, use `wait_agent({ targets: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ target })` each worker.
+
+
+### Model Selection Guide
+
+| Role | model | reasoning_effort | Rationale |
+|------|-------|-------------------|-----------|
+| Scanner (SCAN-*) | (default) | medium | Toolchain scanning + pattern matching, less reasoning |
+| Reviewer (REV-*) | (default) | high | Deep root cause analysis requires full reasoning |
+| Fixer (FIX-*) | (default) | high | Code modification needs precision |
+
+Override model/reasoning_effort in spawn_agent when cost optimization is needed:
+```
+spawn_agent({
+  agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
+  model: "<model-override>",
+  reasoning_effort: "<effort-level>",
+  items: [...]
+})
+```
 
 ## User Commands
 
@@ -121,6 +145,37 @@ After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect
 | `--dimensions=sec,cor,prf,mnt` | Custom dimensions |
 | `-y` / `--yes` | Skip confirmations |
 
+## v4 Agent Coordination
+
+### Message Semantics
+
+| Intent | API | Example |
+|--------|-----|---------|
+| Send scan findings to running reviewer | `send_message` | Queue scan results to REV-* as supplementary context |
+| Not used in this skill | `assign_task` | No resident agents -- sequential 3-stage pipeline |
+| Check running agents | `list_agents` | Verify agent health during resume |
+
+### Pipeline Pattern
+
+This is a **sequential 3-stage pipeline** (scan -> review -> fix). No parallel phases. Each stage completes before the next starts. The coordinator may skip stages (0 findings -> skip review+fix; user declines fix -> skip fix).
+
+### Agent Health Check
+
+Use `list_agents({})` in handleResume and handleComplete:
+
+```
+// Reconcile session state with actual running agents
+const running = list_agents({})
+// Compare with tasks.json active_agents
+// Reset orphaned tasks (in_progress but agent gone) to pending
+```
+
+### Named Agent Targeting
+
+Workers are spawned with `task_name: "<task-id>"` enabling direct addressing:
+- `send_message({ target: "REV-001", items: [...] })` -- queue scan findings to running reviewer
+- `close_agent({ target: "SCAN-001" })` -- cleanup by name after completion
+
 ## Completion Action
 
 When pipeline completes, coordinator presents:

.codex/skills/team-review/roles/coordinator/commands/monitor.md

@@ -54,6 +54,16 @@ Then STOP.
 
 ## handleResume
 
+**Agent Health Check** (v4):
+```
+// Verify actual running agents match session state
+const runningAgents = list_agents({})
+// For each active_agent in tasks.json:
+//   - If agent NOT in runningAgents -> agent crashed
+//   - Reset that task to pending, remove from active_agents
+// This prevents stale agent references from blocking the pipeline
+```
+
 1. Read tasks.json, check active_agents
 2. No active agents -> handleSpawnNext
 3. Has active agents -> check each status
@@ -88,6 +98,7 @@ state.tasks[taskId].status = 'in_progress'
 // 2) Spawn worker
 const agentId = spawn_agent({
   agent_type: "team_worker",
+  task_name: taskId,  // e.g., "SCAN-001" — enables named targeting
   items: [
     { type: "text", text: `## Role Assignment
 role: ${role}
@@ -109,12 +120,18 @@ Execute built-in Phase 1 (task discovery) -> role Phase 2-4 -> built-in Phase 5
 // 3) Track agent
 state.active_agents[taskId] = { agentId, role, started_at: now }
 
-// 4) Wait for completion
-wait_agent({ ids: [agentId] })
-
-// 5) Collect results
-state.tasks[taskId].status = 'completed'
-delete state.active_agents[taskId]
+// 4) Wait for completion — use task_name for stable targeting (v4)
+const waitResult = wait_agent({ targets: [taskId], timeout_ms: 900000 })
+if (waitResult.timed_out) {
+  state.tasks[taskId].status = 'timed_out'
+  close_agent({ target: taskId })
+  delete state.active_agents[taskId]
+} else {
+  // 5) Collect results
+  state.tasks[taskId].status = 'completed'
+  close_agent({ target: taskId })  // Use task_name, not agentId
+  delete state.active_agents[taskId]
+}
 ```
 
    e. Check for checkpoints after worker completes:
@@ -136,8 +153,31 @@ delete state.active_agents[taskId]
 
 5. Update tasks.json, output summary, STOP
 
+**Cross-Agent Supplementary Context** (v4):
+
+When spawning workers in a later pipeline phase, send upstream results as supplementary context to already-running workers:
+
+```
+// Example: Send scan results to running reviewer
+send_message({
+  target: "<running-agent-task-name>",
+  items: [{ type: "text", text: `## Supplementary Context\n${upstreamFindings}` }]
+})
+// Note: send_message queues info without interrupting the agent's current work
+```
+
+Use `send_message` (not `assign_task`) for supplementary info that enriches but doesn't redirect the agent's current task.
+
 ## handleComplete
 
+**Cleanup Verification** (v4):
+```
+// Verify all agents are properly closed
+const remaining = list_agents({})
+// If any team agents still running -> close_agent each
+// Ensures clean session shutdown
+```
+
 Pipeline done. Generate report and completion action.
 
 1. All tasks completed or deleted (no pending, no in_progress)

.codex/skills/team-review/roles/coordinator/role.md

@@ -6,7 +6,7 @@ Orchestrate team-review: parse target -> detect mode -> dispatch task chain -> m
 
 **You are a dispatcher, not a doer.** Your ONLY outputs are:
 - Session state files (`.workflow/.team/` directory)
-- `spawn_agent` / `wait_agent` / `close_agent` / `send_input` calls
+- `spawn_agent` / `wait_agent` / `close_agent` / `send_message` / `assign_task` calls
 - Status reports to the user / `request_user_input` prompts
 
 **FORBIDDEN** (even if the task seems trivial):
@@ -36,6 +36,8 @@ WRONG: Edit/Write on project source files            — worker work
 - Maintain session state (tasks.json)
 - Execute completion action when pipeline finishes
 - **Always proceed through full Phase 1-5 workflow, never skip to direct execution**
+- Use `send_message` for supplementary context (non-interrupting) and `assign_task` for triggering new work
+- Use `list_agents` for session resume health checks and cleanup verification
 
 ### MUST NOT
 - Run analysis tools directly (semgrep, eslint, tsc, etc.)
@@ -151,6 +153,16 @@ Delegate to @commands/monitor.md#handleSpawnNext:
    - auto_archive -> Archive & Clean
    - auto_keep -> Keep Active
 
+## v4 Coordination Patterns
+
+### Message Semantics
+- **send_message**: Queue supplementary info to a running agent. Does NOT interrupt current processing. Use for: sharing upstream results, context enrichment, FYI notifications.
+- **assign_task**: Assign new work and trigger processing. Use for: waking idle agents, redirecting work, requesting new output.
+
+### Agent Lifecycle Management
+- **list_agents({})**: Returns all running agents. Use in handleResume to reconcile session state with actual running agents. Use in handleComplete to verify clean shutdown.
+- **Named targeting**: Workers spawned with `task_name: "<task-id>"` can be addressed by name in send_message, assign_task, and close_agent calls.
+
 ## Error Handling
 
 | Error | Resolution |

.codex/skills/team-roadmap-dev/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: team-roadmap-dev
 description: Unified team skill for roadmap-driven development workflow. Coordinator discusses roadmap with user, then dispatches phased execution pipeline (plan -> execute -> verify). All roles invoke this skill with --role arg. Triggers on "team roadmap-dev".
-allowed-tools: spawn_agent(*), wait_agent(*), send_input(*), close_agent(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*)
+allowed-tools: spawn_agent(*), wait_agent(*), send_message(*), assign_task(*), close_agent(*), list_agents(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*)
 ---
 
 # Team Roadmap Dev
@@ -59,7 +59,8 @@ Before calling ANY tool, apply this check:
 
 | Tool Call | Verdict | Reason |
 |-----------|---------|--------|
-| `spawn_agent`, `wait_agent`, `close_agent`, `send_input` | ALLOWED | Orchestration |
+| `spawn_agent`, `wait_agent`, `close_agent`, `send_message`, `assign_task` | ALLOWED | Orchestration |
+| `list_agents` | ALLOWED | Agent health check |
 | `request_user_input` | ALLOWED | User interaction |
 | `mcp__ccw-tools__team_msg` | ALLOWED | Message bus |
 | `Read/Write` on `.workflow/.team/` files | ALLOWED | Session state |
@@ -90,6 +91,8 @@ Coordinator spawns workers using this template:
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
   items: [
     { type: "text", text: `## Role Assignment
 role: <role>
@@ -113,10 +116,23 @@ pipeline_phase: <pipeline-phase>` },
 })
 ```
 
-After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ id })` each worker.
+After spawning, use `wait_agent({ targets: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ target })` each worker.
 
 **All worker roles** (planner, executor, verifier): Set `inner_loop: true`.
 
+
+### Model Selection Guide
+
+Roadmap development is context-heavy with multi-phase execution. All roles use inner_loop and need high reasoning for complex planning/execution.
+
+| Role | reasoning_effort | Rationale |
+|------|-------------------|-----------|
+| planner | high | Phase planning requires understanding full roadmap context |
+| executor | high | Implementation must align with phase plan precisely |
+| verifier | high | Gap detection requires thorough verification against plan |
+
+All roles are inner_loop=true, enabling coordinator to send additional context via `assign_task` as phases progress.
+
 ## User Commands
 
 | Command | Action |
@@ -150,6 +166,42 @@ After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect
     +-- meta.json               # Session metadata + shared state
 ```
 
+## v4 Agent Coordination
+
+### Message Semantics
+
+| Intent | API | Example |
+|--------|-----|---------|
+| Queue supplementary info (don't interrupt) | `send_message` | Send phase context to running executor |
+| Assign phase work / gap closure | `assign_task` | Assign gap closure iteration to executor after verify |
+| Check running agents | `list_agents` | Verify agent health during resume |
+
+### Agent Health Check
+
+Use `list_agents({})` in handleResume and handleComplete:
+
+```
+// Reconcile session state with actual running agents
+const running = list_agents({})
+// Compare with state.md and config.json active tasks
+// Reset orphaned tasks (in_progress but agent gone) to pending
+```
+
+### Named Agent Targeting
+
+Workers are spawned with `task_name: "<task-id>"` enabling direct addressing:
+- `send_message({ target: "EXEC-N01", items: [...] })` -- send supplementary context to executor
+- `assign_task({ target: "PLAN-N01", items: [...] })` -- assign next phase planning
+- `close_agent({ target: "VERIFY-N01" })` -- cleanup after verification
+
+### Multi-Phase Context Accumulation
+
+Each phase builds on previous phase results. Coordinator accumulates context across phases:
+- Phase N planner receives: roadmap.md + state.md + all previous phase summaries
+- Phase N executor receives: phase plan + previous phase implementation context
+- Phase N verifier receives: phase plan + executor results + success criteria from roadmap
+- On gap closure: verifier findings are sent back to executor via `assign_task` (max 3 iterations)
+
 ## Completion Action
 
 When the pipeline completes:

.codex/skills/team-roadmap-dev/roles/coordinator/commands/monitor.md

@@ -261,6 +261,7 @@ Ready tasks found?
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: taskId,  // e.g., "PLAN-101" — enables named targeting
   items: [{
     description: "Spawn <role> worker for <subject>",
     team_name: "roadmap-dev",
@@ -286,7 +287,22 @@ Execute built-in Phase 1 -> role-spec Phase 2-4 -> built-in Phase 5.`
 })
 ```
 
-Workers report results via `report_agent_job_result()`. Coordinator receives results via `wait_agent({ ids })`.
+Workers report results via `report_agent_job_result()`. Coordinator receives results via `wait_agent({ targets: [taskId], timeout_ms: 900000 })`. If `result.timed_out`, mark tasks as `timed_out` and close agents. Use `close_agent({ target: taskId })` with task_name for cleanup.
+
+**Cross-Agent Supplementary Context** (v4):
+
+When spawning workers in a later pipeline phase, send upstream results as supplementary context to already-running workers:
+
+```
+// Example: Send planning results to running executor
+send_message({
+  target: "<running-agent-task-name>",
+  items: [{ type: "text", text: `## Supplementary Context\n${upstreamFindings}` }]
+})
+// Note: send_message queues info without interrupting the agent's current work
+```
+
+Use `send_message` (not `assign_task`) for supplementary info that enriches but doesn't redirect the agent's current task.
 
 ---
 
@@ -330,6 +346,16 @@ Then STOP.
 
 ### Handler: handleResume
 
+**Agent Health Check** (v4):
+```
+// Verify actual running agents match session state
+const runningAgents = list_agents({})
+// For each active_agent in tasks.json:
+//   - If agent NOT in runningAgents -> agent crashed
+//   - Reset that task to pending, remove from active_agents
+// This prevents stale agent references from blocking the pipeline
+```
+
 Check active worker completion, process results, advance pipeline. Also handles resume from paused state.
 
 ```
@@ -354,6 +380,14 @@ Check coordinates.status:
 
 ### Handler: handleComplete
 
+**Cleanup Verification** (v4):
+```
+// Verify all agents are properly closed
+const remaining = list_agents({})
+// If any team agents still running -> close_agent each
+// Ensures clean session shutdown
+```
+
 All phases done. Generate final project summary and finalize session.
 
 ```

.codex/skills/team-roadmap-dev/roles/coordinator/role.md

@@ -6,7 +6,7 @@ Orchestrate the roadmap-driven development workflow: init prerequisites -> roadm
 
 **You are a dispatcher, not a doer.** Your ONLY outputs are:
 - Session state files (`.workflow/.team/` directory)
-- `spawn_agent` / `wait_agent` / `close_agent` / `send_input` calls
+- `spawn_agent` / `wait_agent` / `close_agent` / `send_message` / `assign_task` calls
 - Status reports to the user / `request_user_input` prompts
 
 **FORBIDDEN** (even if the task seems trivial):
@@ -41,6 +41,8 @@ WRONG: Edit/Write on project source files            — worker work
 - Monitor progress via worker callbacks and route messages
 - Maintain session state persistence
 - **Always proceed through full Phase 1-5 workflow, never skip to direct execution**
+- Use `send_message` for supplementary context (non-interrupting) and `assign_task` for triggering new work
+- Use `list_agents` for session resume health checks and cleanup verification
 
 ### MUST NOT
 
@@ -135,6 +137,7 @@ For callback/check/resume/complete: load `@commands/monitor.md` and execute matc
 | `report_agent_job_result` | Worker communication | workers | Report results to coordinator |
 | `mcp__ccw-tools__team_msg` | Message bus | coordinator | Log all communications |
 | `Read/Write` | File operations | coordinator | Session state management |
+| `list_agents` | System | coordinator | Runtime agent discovery and health check |
 
 ---
 
@@ -262,7 +265,7 @@ mcp__ccw-tools__team_msg({
 
 **Design**: Spawn-and-Stop + Callback pattern.
 - Spawn workers with `spawn_agent()` -> immediately return
-- Worker completes -> report_agent_job_result() -> coordinator receives via wait_agent() -> auto-advance
+- Worker completes -> report_agent_job_result() -> coordinator receives via wait_agent({ timeout_ms: 900000 }) -> auto-advance
 - User can use "check" / "resume" to manually advance
 - Coordinator does one operation per invocation, then STOPS
 
@@ -304,6 +307,16 @@ Delegate to `@commands/monitor.md`:
 
 ---
 
+## v4 Coordination Patterns
+
+### Message Semantics
+- **send_message**: Queue supplementary info to a running agent. Does NOT interrupt current processing. Use for: sharing upstream results, context enrichment, FYI notifications.
+- **assign_task**: Assign new work and trigger processing. Use for: waking idle agents, redirecting work, requesting new output.
+
+### Agent Lifecycle Management
+- **list_agents({})**: Returns all running agents. Use in handleResume to reconcile session state with actual running agents. Use in handleComplete to verify clean shutdown.
+- **Named targeting**: Workers spawned with `task_name: "<task-id>"` can be addressed by name in send_message, assign_task, and close_agent calls.
+
 ## Error Handling
 
 | Scenario | Resolution |

.codex/skills/team-tech-debt/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: team-tech-debt
 description: Unified team skill for tech debt identification and remediation. Scans codebase for tech debt, assesses severity, plans and executes fixes with validation. Uses team-worker agent architecture with roles/ for domain logic. Coordinator orchestrates pipeline, workers are team-worker agents. Triggers on "team tech debt".
-allowed-tools: spawn_agent(*), wait_agent(*), send_input(*), close_agent(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*), mcp__ace-tool__search_context(*), mcp__ccw-tools__read_file(*), mcp__ccw-tools__write_file(*), mcp__ccw-tools__edit_file(*), mcp__ccw-tools__team_msg(*)
+allowed-tools: spawn_agent(*), wait_agent(*), send_message(*), assign_task(*), close_agent(*), list_agents(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*), mcp__ace-tool__search_context(*), mcp__ccw-tools__read_file(*), mcp__ccw-tools__write_file(*), mcp__ccw-tools__edit_file(*), mcp__ccw-tools__team_msg(*)
 ---
 
 # Team Tech Debt
@@ -55,7 +55,8 @@ Before calling ANY tool, apply this check:
 
 | Tool Call | Verdict | Reason |
 |-----------|---------|--------|
-| `spawn_agent`, `wait_agent`, `close_agent`, `send_input` | ALLOWED | Orchestration |
+| `spawn_agent`, `wait_agent`, `close_agent`, `send_message`, `assign_task` | ALLOWED | Orchestration |
+| `list_agents` | ALLOWED | Agent health check |
 | `request_user_input` | ALLOWED | User interaction |
 | `mcp__ccw-tools__team_msg` | ALLOWED | Message bus |
 | `Read/Write` on `.workflow/.team/` files | ALLOWED | Session state |
@@ -86,6 +87,8 @@ Coordinator spawns workers using this template:
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
   items: [
     { type: "text", text: `## Role Assignment
 role: <role>
@@ -109,7 +112,29 @@ pipeline_phase: <pipeline-phase>` },
 })
 ```
 
-After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ id })` each worker.
+After spawning, use `wait_agent({ targets: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ target })` each worker.
+
+
+### Model Selection Guide
+
+Tech debt follows a discovery-to-fix pipeline. Scanner is broad/fast, later stages need deeper reasoning.
+
+| Role | reasoning_effort | Rationale |
+|------|-------------------|-----------|
+| scanner | medium | Broad codebase scan, pattern matching over deep analysis |
+| assessor | high | Severity assessment requires understanding impact and risk |
+| planner | high | Remediation planning must prioritize and sequence fixes |
+| executor | high | Code fixes must preserve behavior while removing debt |
+| validator | medium | Validation follows defined acceptance criteria |
+
+### Pipeline Pattern: Scanner Results Inform Downstream
+
+Scanner discoveries flow through the pipeline — each stage narrows and refines:
+1. TDSCAN produces broad debt inventory
+2. TDEVAL assesses and prioritizes (filters low-impact items)
+3. TDPLAN creates sequenced fix plan from assessed items
+4. TDFIX implements fixes per plan
+5. TDVAL validates fixes against original debt findings
 
 ## User Commands
 
@@ -141,6 +166,34 @@ After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect
 └── wisdom/                 # Cross-task knowledge
 ```
 
+## v4 Agent Coordination
+
+### Message Semantics
+
+| Intent | API | Example |
+|--------|-----|---------|
+| Queue supplementary info (don't interrupt) | `send_message` | Send scan findings to running assessor |
+| Assign fix from remediation plan | `assign_task` | Assign TDFIX task from planner output |
+| Check running agents | `list_agents` | Verify agent health during resume |
+
+### Agent Health Check
+
+Use `list_agents({})` in handleResume and handleComplete:
+
+```
+// Reconcile session state with actual running agents
+const running = list_agents({})
+// Compare with meta.json active tasks
+// Reset orphaned tasks (in_progress but agent gone) to pending
+```
+
+### Named Agent Targeting
+
+Workers are spawned with `task_name: "<task-id>"` enabling direct addressing:
+- `send_message({ target: "TDSCAN-001", items: [...] })` -- send additional scan scope to scanner
+- `assign_task({ target: "TDFIX-001", items: [...] })` -- assign fix task from planner output
+- `close_agent({ target: "TDVAL-001" })` -- cleanup after validation
+
 ## Error Handling
 
 | Scenario | Resolution |

.codex/skills/team-tech-debt/roles/coordinator/commands/monitor.md

@@ -43,6 +43,16 @@ Output status -- do NOT advance pipeline.
 
 ## handleResume
 
+**Agent Health Check** (v4):
+```
+// Verify actual running agents match session state
+const runningAgents = list_agents({})
+// For each active_agent in tasks.json:
+//   - If agent NOT in runningAgents -> agent crashed
+//   - Reset that task to pending, remove from active_agents
+// This prevents stale agent references from blocking the pipeline
+```
+
 1. Read tasks.json, check active_agents
 2. Tasks stuck in "in_progress" -> reset to "pending"
 3. Tasks with completed deps but still "pending" -> include in spawn list
@@ -72,6 +82,7 @@ state.tasks[taskId].status = 'in_progress'
 // 2) Spawn worker
 const agentId = spawn_agent({
   agent_type: "team_worker",
+  task_name: taskId,  // e.g., "TDSCAN-001" — enables named targeting
   items: [
     { type: "text", text: `## Role Assignment
 role: ${task.role}
@@ -112,15 +123,22 @@ Stage-to-role mapping:
 After spawning all ready tasks:
 
 ```javascript
-// 4) Batch wait for all spawned workers
-const agentIds = Object.values(state.active_agents).map(a => a.agentId)
-wait_agent({ ids: agentIds, timeout_ms: 900000 })
-
-// 5) Collect results
-for (const [taskId, agent] of Object.entries(state.active_agents)) {
-  state.tasks[taskId].status = 'completed'
-  close_agent({ id: agent.agentId })
-  delete state.active_agents[taskId]
+// 4) Batch wait — use task_name for stable targeting (v4)
+const taskNames = Object.keys(state.active_agents)
+const waitResult = wait_agent({ targets: taskNames, timeout_ms: 900000 })
+if (waitResult.timed_out) {
+  for (const taskId of taskNames) {
+    state.tasks[taskId].status = 'timed_out'
+    close_agent({ target: taskId })
+    delete state.active_agents[taskId]
+  }
+} else {
+  // 5) Collect results
+  for (const [taskId, agent] of Object.entries(state.active_agents)) {
+    state.tasks[taskId].status = 'completed'
+    close_agent({ target: taskId })  // Use task_name, not agentId
+    delete state.active_agents[taskId]
+  }
 }
 ```
 
@@ -188,6 +206,21 @@ After task completion, check for checkpoints:
   }
   ```
 
+**Cross-Agent Supplementary Context** (v4):
+
+When spawning workers in a later pipeline phase, send upstream results as supplementary context to already-running workers:
+
+```
+// Example: Send scan results to running assessor
+send_message({
+  target: "<running-agent-task-name>",
+  items: [{ type: "text", text: `## Supplementary Context\n${upstreamFindings}` }]
+})
+// Note: send_message queues info without interrupting the agent's current work
+```
+
+Use `send_message` (not `assign_task`) for supplementary info that enriches but doesn't redirect the agent's current task.
+
 ### Persist and Loop
 
 After processing all results:
@@ -199,6 +232,14 @@ After processing all results:
 
 ## handleComplete
 
+**Cleanup Verification** (v4):
+```
+// Verify all agents are properly closed
+const remaining = list_agents({})
+// If any team agents still running -> close_agent each
+// Ensures clean session shutdown
+```
+
 Pipeline done. Generate report and completion action.
 
 1. Verify all tasks (including fix-verify tasks) have status "completed"

.codex/skills/team-tech-debt/roles/coordinator/role.md

@@ -6,7 +6,7 @@
 
 **You are a dispatcher, not a doer.** Your ONLY outputs are:
 - Session state files (`.workflow/.team/` directory)
-- `spawn_agent` / `wait_agent` / `close_agent` / `send_input` calls
+- `spawn_agent` / `wait_agent` / `close_agent` / `send_message` / `assign_task` calls
 - Status reports to the user / `request_user_input` prompts
 
 **FORBIDDEN** (even if the task seems trivial):
@@ -33,6 +33,8 @@ WRONG: Edit/Write on project source files            — worker work
 - Monitor worker progress via spawn_agent/wait_agent and route messages
 - Maintain session state persistence (tasks.json)
 - **Always proceed through full Phase 1-5 workflow, never skip to direct execution**
+- Use `send_message` for supplementary context (non-interrupting) and `assign_task` for triggering new work
+- Use `list_agents` for session resume health checks and cleanup verification
 
 ### MUST NOT
 - Execute tech debt work directly (delegate to workers)
@@ -147,6 +149,16 @@ Delegate to @commands/monitor.md#handleSpawnNext:
 5. Output with [coordinator] prefix
 6. Execute completion action (request_user_input: 新目标 / 深度修复 / 关闭团队)
 
+## v4 Coordination Patterns
+
+### Message Semantics
+- **send_message**: Queue supplementary info to a running agent. Does NOT interrupt current processing. Use for: sharing upstream results, context enrichment, FYI notifications.
+- **assign_task**: Assign new work and trigger processing. Use for: waking idle agents, redirecting work, requesting new output.
+
+### Agent Lifecycle Management
+- **list_agents({})**: Returns all running agents. Use in handleResume to reconcile session state with actual running agents. Use in handleComplete to verify clean shutdown.
+- **Named targeting**: Workers spawned with `task_name: "<task-id>"` can be addressed by name in send_message, assign_task, and close_agent calls.
+
 ## Error Handling
 
 | Error | Resolution |

.codex/skills/team-testing/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: team-testing
 description: Unified team skill for testing team. Progressive test coverage through Generator-Critic loops, shared memory, and dynamic layer selection. Triggers on "team testing".
-allowed-tools: spawn_agent(*), wait_agent(*), send_input(*), close_agent(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*)
+allowed-tools: spawn_agent(*), wait_agent(*), send_message(*), assign_task(*), close_agent(*), list_agents(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*)
 ---
 
 # Team Testing
@@ -54,7 +54,8 @@ Before calling ANY tool, apply this check:
 
 | Tool Call | Verdict | Reason |
 |-----------|---------|--------|
-| `spawn_agent`, `wait_agent`, `close_agent`, `send_input` | ALLOWED | Orchestration |
+| `spawn_agent`, `wait_agent`, `close_agent`, `send_message`, `assign_task` | ALLOWED | Orchestration |
+| `list_agents` | ALLOWED | Agent health check |
 | `request_user_input` | ALLOWED | User interaction |
 | `mcp__ccw-tools__team_msg` | ALLOWED | Message bus |
 | `Read/Write` on `.workflow/.team/` files | ALLOWED | Session state |
@@ -85,6 +86,8 @@ Coordinator spawns workers using this template:
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
   items: [
     { type: "text", text: `## Role Assignment
 role: <role>
@@ -108,7 +111,29 @@ pipeline_phase: <pipeline-phase>` },
 })
 ```
 
-After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ id })` each worker.
+After spawning, use `wait_agent({ targets: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ target })` each worker.
+
+
+### Model Selection Guide
+
+| Role | model | reasoning_effort | Rationale |
+|------|-------|-------------------|-----------|
+| Strategist (STRATEGY-*) | (default) | high | Test strategy requires deep code understanding |
+| Generator (TESTGEN-*) | (default) | high | Test code generation needs precision |
+| Executor (TESTRUN-*) | (default) | medium | Running tests and collecting results, less reasoning |
+| Analyst (TESTANA-*) | (default) | high | Coverage analysis and quality assessment |
+
+Override model/reasoning_effort in spawn_agent when cost optimization is needed:
+```
+spawn_agent({
+  agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
+  model: "<model-override>",
+  reasoning_effort: "<effort-level>",
+  items: [...]
+})
+```
 
 ## User Commands
 
@@ -119,6 +144,50 @@ After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect
 | `revise <TASK-ID>` | Revise specific task |
 | `feedback <text>` | Inject feedback for revision |
 
+## v4 Agent Coordination
+
+### Message Semantics
+
+| Intent | API | Example |
+|--------|-----|---------|
+| Send strategy to running generators | `send_message` | Queue test strategy findings to TESTGEN-* workers |
+| Not used in this skill | `assign_task` | No resident agents -- all workers are one-shot |
+| Check running agents | `list_agents` | Verify parallel generator/executor health |
+
+### Parallel Test Generation
+
+Comprehensive pipeline spawns multiple generators (per layer) and executors in parallel:
+
+```
+// Spawn parallel generators for L1 and L2
+const genNames = ["TESTGEN-001", "TESTGEN-002"]
+for (const name of genNames) {
+  spawn_agent({ agent_type: "team_worker", task_name: name, ... })
+}
+wait_agent({ targets: genNames, timeout_ms: 900000 })
+```
+
+### GC Loop Coordination
+
+Generator-Critic loops create dynamic TESTGEN-fix and TESTRUN-fix tasks. The coordinator tracks `gc_rounds[layer]` and creates fix tasks dynamically when coverage is below target.
+
+### Agent Health Check
+
+Use `list_agents({})` in handleResume and handleComplete:
+
+```
+// Reconcile session state with actual running agents
+const running = list_agents({})
+// Compare with tasks.json active_agents
+// Reset orphaned tasks (in_progress but agent gone) to pending
+```
+
+### Named Agent Targeting
+
+Workers are spawned with `task_name: "<task-id>"` enabling direct addressing:
+- `send_message({ target: "TESTGEN-001", items: [...] })` -- queue strategy context to running generator
+- `close_agent({ target: "TESTRUN-001" })` -- cleanup by name after wait_agent returns
+
 ## Completion Action
 
 When pipeline completes, coordinator presents:

.codex/skills/team-testing/roles/coordinator/commands/monitor.md

@@ -59,6 +59,16 @@ Then STOP.
 
 ## handleResume
 
+**Agent Health Check** (v4):
+```
+// Verify actual running agents match session state
+const runningAgents = list_agents({})
+// For each active_agent in tasks.json:
+//   - If agent NOT in runningAgents -> agent crashed
+//   - Reset that task to pending, remove from active_agents
+// This prevents stale agent references from blocking the pipeline
+```
+
 1. Read tasks.json, check active_agents
 2. No active agents -> handleSpawnNext
 3. Has active agents -> check each status
@@ -96,6 +106,7 @@ state.tasks[taskId].status = 'in_progress'
 // 2) Spawn worker
 const agentId = spawn_agent({
   agent_type: "team_worker",
+  task_name: taskId,  // e.g., "STRATEGY-001" — enables named targeting
   items: [
     { type: "text", text: `## Role Assignment
 role: ${task.role}
@@ -135,15 +146,22 @@ state.active_agents[taskId] = { agentId, role: task.role, started_at: now }
 After spawning all ready tasks:
 
 ```javascript
-// 4) Batch wait for all spawned workers
-const agentIds = Object.values(state.active_agents).map(a => a.agentId)
-wait_agent({ ids: agentIds, timeout_ms: 900000 })
-
-// 5) Collect results
-for (const [taskId, agent] of Object.entries(state.active_agents)) {
-  state.tasks[taskId].status = 'completed'
-  close_agent({ id: agent.agentId })
-  delete state.active_agents[taskId]
+// 4) Batch wait — use task_name for stable targeting (v4)
+const taskNames = Object.keys(state.active_agents)
+const waitResult = wait_agent({ targets: taskNames, timeout_ms: 900000 })
+if (waitResult.timed_out) {
+  for (const taskId of taskNames) {
+    state.tasks[taskId].status = 'timed_out'
+    close_agent({ target: taskId })
+    delete state.active_agents[taskId]
+  }
+} else {
+  // 5) Collect results
+  for (const [taskId, agent] of Object.entries(state.active_agents)) {
+    state.tasks[taskId].status = 'completed'
+    close_agent({ target: taskId })  // Use task_name, not agentId
+    delete state.active_agents[taskId]
+  }
 }
 ```
 
@@ -182,6 +200,21 @@ Add to tasks.json:
 ```
 Update tasks.json gc_rounds[layer]++
 
+**Cross-Agent Supplementary Context** (v4):
+
+When spawning workers in a later pipeline phase, send upstream results as supplementary context to already-running workers:
+
+```
+// Example: Send strategy results to running generators
+send_message({
+  target: "<running-agent-task-name>",
+  items: [{ type: "text", text: `## Supplementary Context\n${upstreamFindings}` }]
+})
+// Note: send_message queues info without interrupting the agent's current work
+```
+
+Use `send_message` (not `assign_task`) for supplementary info that enriches but doesn't redirect the agent's current task.
+
 ### Persist and Loop
 
 After processing all results:
@@ -193,6 +226,14 @@ After processing all results:
 
 ## handleComplete
 
+**Cleanup Verification** (v4):
+```
+// Verify all agents are properly closed
+const remaining = list_agents({})
+// If any team agents still running -> close_agent each
+// Ensures clean session shutdown
+```
+
 Pipeline done. Generate report and completion action.
 
 1. Verify all tasks (including any GC fix tasks) have status "completed" or "failed"

.codex/skills/team-testing/roles/coordinator/role.md

@@ -6,7 +6,7 @@ Orchestrate team-testing: analyze -> dispatch -> spawn -> monitor -> report.
 
 **You are a dispatcher, not a doer.** Your ONLY outputs are:
 - Session state files (`.workflow/.team/` directory)
-- `spawn_agent` / `wait_agent` / `close_agent` / `send_input` calls
+- `spawn_agent` / `wait_agent` / `close_agent` / `send_message` / `assign_task` calls
 - Status reports to the user / `request_user_input` prompts
 
 **FORBIDDEN** (even if the task seems trivial):
@@ -34,6 +34,8 @@ WRONG: Edit/Write on test or source files            — worker work
 - Handle Generator-Critic cycles with max 3 iterations per layer
 - Execute completion action in Phase 5
 - **Always proceed through full Phase 1-5 workflow, never skip to direct execution**
+- Use `send_message` for supplementary context (non-interrupting) and `assign_task` for triggering new work
+- Use `list_agents` for session resume health checks and cleanup verification
 
 ### MUST NOT
 - Implement domain logic (test generation, execution, analysis) -- workers handle this
@@ -160,6 +162,16 @@ Delegate to @commands/monitor.md#handleSpawnNext:
    - auto_archive -> Archive & Clean (rm -rf session folder)
    - auto_keep -> Keep Active
 
+## v4 Coordination Patterns
+
+### Message Semantics
+- **send_message**: Queue supplementary info to a running agent. Does NOT interrupt current processing. Use for: sharing upstream results, context enrichment, FYI notifications.
+- **assign_task**: Assign new work and trigger processing. Use for: waking idle agents, redirecting work, requesting new output.
+
+### Agent Lifecycle Management
+- **list_agents({})**: Returns all running agents. Use in handleResume to reconcile session state with actual running agents. Use in handleComplete to verify clean shutdown.
+- **Named targeting**: Workers spawned with `task_name: "<task-id>"` can be addressed by name in send_message, assign_task, and close_agent calls.
+
 ## Error Handling
 
 | Error | Resolution |

.codex/skills/team-uidesign/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: team-uidesign
 description: Unified team skill for UI design team. Research -> design tokens -> audit -> implementation. Uses team-worker agent architecture with roles/ for domain logic. Coordinator orchestrates dual-track pipeline with GC loops and sync points. Triggers on "team ui design", "ui design team".
-allowed-tools: spawn_agent(*), wait_agent(*), send_input(*), close_agent(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*), mcp__ace-tool__search_context(*), mcp__ccw-tools__read_file(*), mcp__ccw-tools__write_file(*), mcp__ccw-tools__edit_file(*), mcp__ccw-tools__team_msg(*)
+allowed-tools: spawn_agent(*), wait_agent(*), send_message(*), assign_task(*), close_agent(*), list_agents(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*), mcp__ace-tool__search_context(*), mcp__ccw-tools__read_file(*), mcp__ccw-tools__write_file(*), mcp__ccw-tools__edit_file(*), mcp__ccw-tools__team_msg(*)
 ---
 
 # Team UI Design
@@ -54,7 +54,8 @@ Before calling ANY tool, apply this check:
 
 | Tool Call | Verdict | Reason |
 |-----------|---------|--------|
-| `spawn_agent`, `wait_agent`, `close_agent`, `send_input` | ALLOWED | Orchestration |
+| `spawn_agent`, `wait_agent`, `close_agent`, `send_message`, `assign_task` | ALLOWED | Orchestration |
+| `list_agents` | ALLOWED | Agent health check |
 | `request_user_input` | ALLOWED | User interaction |
 | `mcp__ccw-tools__team_msg` | ALLOWED | Message bus |
 | `Read/Write` on `.workflow/.team/` files | ALLOWED | Session state |
@@ -85,6 +86,8 @@ Coordinator spawns workers using this template:
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
   items: [
     { type: "text", text: `## Role Assignment
 role: <role>
@@ -108,7 +111,38 @@ pipeline_phase: <pipeline-phase>` },
 })
 ```
 
-After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ id })` each worker.
+After spawning, use `wait_agent({ targets: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ target })` each worker.
+
+
+### Model Selection Guide
+
+UI design is a creative pipeline where research findings inform design decisions. Researcher needs thorough analysis, designer needs creative reasoning.
+
+| Role | reasoning_effort | Rationale |
+|------|-------------------|-----------|
+| researcher | high | Deep analysis of existing design systems and patterns |
+| designer | high | Creative design token/component spec generation |
+| reviewer | high | Design audit must catch consistency and accessibility issues |
+| implementer | medium | Implementation follows defined design specs |
+
+### Research-to-Design Context Flow
+
+Researcher findings must reach designer via coordinator's upstream context:
+```
+// After RESEARCH-001 completes, coordinator sends findings to designer
+spawn_agent({
+  agent_type: "team_worker",
+  task_name: "DESIGN-001",
+  fork_context: false,
+  items: [
+    ...,
+    { type: "text", text: `## Upstream Context
+Research findings: <session>/research/design-intelligence.json
+Component inventory: <session>/research/component-inventory.json
+Accessibility audit: <session>/research/accessibility-audit.json` }
+  ]
+})
+```
 
 ## User Commands
 
@@ -145,6 +179,34 @@ After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect
 └── wisdom/                    # Cross-task knowledge
 ```
 
+## v4 Agent Coordination
+
+### Message Semantics
+
+| Intent | API | Example |
+|--------|-----|---------|
+| Queue supplementary info (don't interrupt) | `send_message` | Send research findings to running designer |
+| Assign build from reviewed specs | `assign_task` | Assign BUILD task after audit passes |
+| Check running agents | `list_agents` | Verify agent health during resume |
+
+### Agent Health Check
+
+Use `list_agents({})` in handleResume and handleComplete:
+
+```
+// Reconcile session state with actual running agents
+const running = list_agents({})
+// Compare with meta.json active tasks
+// Reset orphaned tasks (in_progress but agent gone) to pending
+```
+
+### Named Agent Targeting
+
+Workers are spawned with `task_name: "<task-id>"` enabling direct addressing:
+- `send_message({ target: "DESIGN-001", items: [...] })` -- send additional research findings to designer
+- `assign_task({ target: "BUILD-001", items: [...] })` -- assign implementation from reviewed design specs
+- `close_agent({ target: "AUDIT-001" })` -- cleanup after design audit
+
 ## Error Handling
 
 | Scenario | Resolution |

.codex/skills/team-uidesign/roles/coordinator/commands/monitor.md

@@ -99,6 +99,16 @@ Output status -- do NOT advance pipeline.
 
 ## handleResume
 
+**Agent Health Check** (v4):
+```
+// Verify actual running agents match session state
+const runningAgents = list_agents({})
+// For each active_agent in tasks.json:
+//   - If agent NOT in runningAgents -> agent crashed
+//   - Reset that task to pending, remove from active_agents
+// This prevents stale agent references from blocking the pipeline
+```
+
 1. Audit tasks.json for inconsistencies:
    - Tasks stuck in "in_progress" -> reset to "pending"
    - Tasks with completed deps but still "pending" -> include in spawn list
@@ -124,6 +134,7 @@ state.tasks[taskId].status = 'in_progress'
 // 2) Spawn worker
 const agentId = spawn_agent({
   agent_type: "team_worker",
+  task_name: taskId,  // e.g., "DESIGN-001" — enables named targeting
   items: [
     { type: "text", text: `## Role Assignment
 role: ${role}
@@ -141,12 +152,18 @@ Execute built-in Phase 1 (task discovery) -> role Phase 2-4 -> built-in Phase 5
 // 3) Track agent
 state.active_agents[taskId] = { agentId, role, started_at: now }
 
-// 4) Wait for completion
-wait_agent({ ids: [agentId] })
-
-// 5) Collect results and update tasks.json
-state.tasks[taskId].status = 'completed'
-delete state.active_agents[taskId]
+// 4) Wait for completion — use task_name for stable targeting (v4)
+const waitResult = wait_agent({ targets: [taskId], timeout_ms: 900000 })
+if (waitResult.timed_out) {
+  state.tasks[taskId].status = 'timed_out'
+  close_agent({ target: taskId })
+  delete state.active_agents[taskId]
+} else {
+  // 5) Collect results and update tasks.json
+  state.tasks[taskId].status = 'completed'
+  close_agent({ target: taskId })  // Use task_name, not agentId
+  delete state.active_agents[taskId]
+}
 ```
 
 **Parallel spawn rules by mode**:
@@ -159,10 +176,33 @@ delete state.active_agents[taskId]
 | full-system | After Sync Point 1 | Spawn DESIGN-002 + BUILD-001 in parallel, wait_agent for both |
 | full-system | After BUILD-002 | Spawn AUDIT-003 |
 
+**Cross-Agent Supplementary Context** (v4):
+
+When spawning workers in a later pipeline phase, send upstream results as supplementary context to already-running workers:
+
+```
+// Example: Send design results to running implementer
+send_message({
+  target: "<running-agent-task-name>",
+  items: [{ type: "text", text: `## Supplementary Context\n${upstreamFindings}` }]
+})
+// Note: send_message queues info without interrupting the agent's current work
+```
+
+Use `send_message` (not `assign_task`) for supplementary info that enriches but doesn't redirect the agent's current task.
+
 5. Update tasks.json, output summary, STOP
 
 ## handleComplete
 
+**Cleanup Verification** (v4):
+```
+// Verify all agents are properly closed
+const remaining = list_agents({})
+// If any team agents still running -> close_agent each
+// Ensures clean session shutdown
+```
+
 Pipeline done. Generate report and completion action.
 
 **Completion check by mode**:

.codex/skills/team-uidesign/roles/coordinator/role.md

@@ -6,7 +6,7 @@ UI Design Team coordinator. Orchestrate pipeline: analyze -> dispatch -> spawn -
 
 **You are a dispatcher, not a doer.** Your ONLY outputs are:
 - Session state files (`.workflow/.team/` directory)
-- `spawn_agent` / `wait_agent` / `close_agent` / `send_input` calls
+- `spawn_agent` / `wait_agent` / `close_agent` / `send_message` / `assign_task` calls
 - Status reports to the user / `request_user_input` prompts
 
 **FORBIDDEN** (even if the task seems trivial):
@@ -34,6 +34,8 @@ WRONG: Edit/Write on project source files            — worker work
 - Handle Generator-Critic loops with max 2 iterations
 - Maintain session state persistence (tasks.json)
 - **Always proceed through full Phase 1-5 workflow, never skip to direct execution**
+- Use `send_message` for supplementary context (non-interrupting) and `assign_task` for triggering new work
+- Use `list_agents` for session resume health checks and cleanup verification
 
 ### MUST NOT
 - Implement domain logic (researching, designing, auditing, building) -- workers handle this
@@ -187,6 +189,16 @@ Delegate to `@commands/monitor.md#handleSpawnNext`:
    })
    ```
 
+## v4 Coordination Patterns
+
+### Message Semantics
+- **send_message**: Queue supplementary info to a running agent. Does NOT interrupt current processing. Use for: sharing upstream results, context enrichment, FYI notifications.
+- **assign_task**: Assign new work and trigger processing. Use for: waking idle agents, redirecting work, requesting new output.
+
+### Agent Lifecycle Management
+- **list_agents({})**: Returns all running agents. Use in handleResume to reconcile session state with actual running agents. Use in handleComplete to verify clean shutdown.
+- **Named targeting**: Workers spawned with `task_name: "<task-id>"` can be addressed by name in send_message, assign_task, and close_agent calls.
+
 ## Error Handling
 
 | Error | Resolution |

.codex/skills/team-ultra-analyze/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: team-ultra-analyze
 description: Deep collaborative analysis team skill. All roles route via this SKILL.md. Beat model is coordinator-only (monitor.md). Structure is roles/ + specs/. Triggers on "team ultra-analyze", "team analyze".
-allowed-tools: spawn_agent(*), wait_agent(*), send_input(*), close_agent(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*)
+allowed-tools: spawn_agent(*), wait_agent(*), send_message(*), assign_task(*), close_agent(*), list_agents(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*)
 ---
 
 # Team Ultra Analyze
@@ -62,7 +62,8 @@ Before calling ANY tool, apply this check:
 
 | Tool Call | Verdict | Reason |
 |-----------|---------|--------|
-| `spawn_agent`, `wait_agent`, `close_agent`, `send_input` | ALLOWED | Orchestration |
+| `spawn_agent`, `wait_agent`, `close_agent`, `send_message`, `assign_task` | ALLOWED | Orchestration |
+| `list_agents` | ALLOWED | Agent health check |
 | `request_user_input` | ALLOWED | User interaction |
 | `mcp__ccw-tools__team_msg` | ALLOWED | Message bus |
 | `Read/Write` on `.workflow/.team/` files | ALLOWED | Session state |
@@ -93,6 +94,8 @@ Coordinator spawns workers using this template:
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
   items: [
     { type: "text", text: `## Role Assignment
 role: <role>
@@ -116,7 +119,29 @@ pipeline_phase: <pipeline-phase>` },
 })
 ```
 
-After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ id })` each worker.
+After spawning, use `wait_agent({ targets: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ target })` each worker.
+
+
+### Model Selection Guide
+
+| Role | model | reasoning_effort | Rationale |
+|------|-------|-------------------|-----------|
+| Explorer (EXPLORE-*) | (default) | medium | File reading and pattern scanning, less reasoning needed |
+| Analyst (ANALYZE-*) | (default) | high | Deep analysis requires full reasoning |
+| Discussant (DISCUSS-*) | (default) | high | Synthesis of multiple viewpoints, dialectic reasoning |
+| Synthesizer (SYNTH-*) | (default) | medium | Aggregation and summary over generation |
+
+Override model/reasoning_effort in spawn_agent when cost optimization is needed:
+```
+spawn_agent({
+  agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
+  model: "<model-override>",
+  reasoning_effort: "<effort-level>",
+  items: [...]
+})
+```
 
 ## User Commands
 
@@ -148,6 +173,51 @@ After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect
 |   +-- issues.md
 ```
 
+## v4 Agent Coordination
+
+### Message Semantics
+
+| Intent | API | Example |
+|--------|-----|---------|
+| Send exploration findings to running analysts | `send_message` | Queue upstream context without interrupting ANALYZE-* workers |
+| Not used in this skill | `assign_task` | No resident agents -- all workers are one-shot |
+| Check running agents | `list_agents` | Verify parallel explorer/analyst health during resume |
+
+### Parallel Phase Coordination
+
+Standard/Deep modes spawn N parallel agents in EXPLORE and ANALYZE phases. Use batch spawn + wait:
+
+```
+// EXPLORE phase: spawn N explorers in parallel
+const explorerNames = ["EXPLORE-001", "EXPLORE-002", ..., "EXPLORE-00N"]
+for (const name of explorerNames) {
+  spawn_agent({ agent_type: "team_worker", task_name: name, ... })
+}
+wait_agent({ targets: explorerNames, timeout_ms: 900000 })
+// Collect all results, then spawn ANALYZE phase
+
+// ANALYZE phase: send exploration context to analysts via items (not send_message)
+// since analysts are spawned AFTER explorers complete
+```
+
+### Agent Health Check
+
+Use `list_agents({})` in handleResume and handleComplete:
+
+```
+// Reconcile session state with actual running agents
+const running = list_agents({})
+// Compare with tasks.json active_agents
+// Reset orphaned tasks (in_progress but agent gone) to pending
+// Critical for parallel phases -- multiple agents may crash independently
+```
+
+### Named Agent Targeting
+
+Workers are spawned with `task_name: "<task-id>"` enabling direct addressing:
+- `send_message({ target: "ANALYZE-001", items: [...] })` -- queue supplementary exploration findings
+- `close_agent({ target: "EXPLORE-001" })` -- cleanup by name after wait_agent returns
+
 ## Completion Action
 
 When pipeline completes, coordinator presents:

.codex/skills/team-ultra-analyze/roles/coordinator/commands/monitor.md

@@ -210,6 +210,7 @@ state.tasks[taskId].status = 'in_progress'
 // 2) Spawn worker
 const agentId = spawn_agent({
   agent_type: "team_worker",
+  task_name: taskId,  // e.g., "EXPLORE-001" — enables named targeting
   items: [
     { type: "text", text: `## Role Assignment
 role: ${role}
@@ -236,15 +237,22 @@ state.active_agents[taskId] = { agentId, role, started_at: now }
 After spawning all ready tasks:
 
 ```javascript
-// 4) Batch wait for all spawned workers
-const agentIds = Object.values(state.active_agents)
-  .map(a => a.agentId)
-wait_agent({ ids: agentIds })
-
-// 5) Collect results and update tasks.json
-for (const [taskId, agent] of Object.entries(state.active_agents)) {
-  state.tasks[taskId].status = 'completed'
-  delete state.active_agents[taskId]
+// 4) Batch wait — use task_name for stable targeting (v4)
+const taskNames = Object.keys(state.active_agents)
+const waitResult = wait_agent({ targets: taskNames, timeout_ms: 900000 })
+if (waitResult.timed_out) {
+  for (const taskId of taskNames) {
+    state.tasks[taskId].status = 'timed_out'
+    close_agent({ target: taskId })
+    delete state.active_agents[taskId]
+  }
+} else {
+  // 5) Collect results and update tasks.json
+  for (const [taskId, agent] of Object.entries(state.active_agents)) {
+    state.tasks[taskId].status = 'completed'
+    close_agent({ target: taskId })  // Use task_name, not agentId
+    delete state.active_agents[taskId]
+  }
 }
 ```
 
@@ -258,6 +266,21 @@ for (const [taskId, agent] of Object.entries(state.active_agents)) {
 | all | DISCUSS phase | One discussant at a time |
 | all | SYNTH phase | One synthesizer |
 
+**Cross-Agent Supplementary Context** (v4):
+
+When spawning workers in a later pipeline phase, send upstream results as supplementary context to already-running workers:
+
+```
+// Example: Send exploration results to running analysts
+send_message({
+  target: "<running-agent-task-name>",
+  items: [{ type: "text", text: `## Supplementary Context\n${upstreamFindings}` }]
+})
+// Note: send_message queues info without interrupting the agent's current work
+```
+
+Use `send_message` (not `assign_task`) for supplementary info that enriches but doesn't redirect the agent's current task.
+
 5. **STOP** after processing -- wait for next event
 
 ### handleCheck
@@ -282,6 +305,16 @@ Output status -- do NOT advance pipeline.
 
 ### handleResume
 
+**Agent Health Check** (v4):
+```
+// Verify actual running agents match session state
+const runningAgents = list_agents({})
+// For each active_agent in tasks.json:
+//   - If agent NOT in runningAgents -> agent crashed
+//   - Reset that task to pending, remove from active_agents
+// This prevents stale agent references from blocking the pipeline
+```
+
 Resume pipeline after user pause or interruption.
 
 1. Audit tasks.json for inconsistencies:
@@ -291,6 +324,14 @@ Resume pipeline after user pause or interruption.
 
 ### handleComplete
 
+**Cleanup Verification** (v4):
+```
+// Verify all agents are properly closed
+const remaining = list_agents({})
+// If any team agents still running -> close_agent each
+// Ensures clean session shutdown
+```
+
 Triggered when all pipeline tasks are completed.
 
 **Completion check**:

.codex/skills/team-ultra-analyze/roles/coordinator/role.md

@@ -10,7 +10,7 @@ Orchestrates the analysis pipeline: topic clarification, pipeline mode selection
 
 **You are a dispatcher, not a doer.** Your ONLY outputs are:
 - Session state files (`.workflow/.team/` directory)
-- `spawn_agent` / `wait_agent` / `close_agent` / `send_input` calls
+- `spawn_agent` / `wait_agent` / `close_agent` / `send_message` / `assign_task` calls
 - Status reports to the user / `request_user_input` prompts
 
 **FORBIDDEN** (even if the task seems trivial):
@@ -35,6 +35,8 @@ WRONG: Edit/Write on project source files            — worker work
 - Handle discussion loop with max 5 rounds (Deep mode)
 - Execute completion action in Phase 5
 - **Always proceed through full Phase 1-5 workflow, never skip to direct execution**
+- Use `send_message` for supplementary context (non-interrupting) and `assign_task` for triggering new work
+- Use `list_agents` for session resume health checks and cleanup verification
 
 ### MUST NOT
 
@@ -232,6 +234,16 @@ request_user_input({
 
 ---
 
+## v4 Coordination Patterns
+
+### Message Semantics
+- **send_message**: Queue supplementary info to a running agent. Does NOT interrupt current processing. Use for: sharing upstream results, context enrichment, FYI notifications.
+- **assign_task**: Assign new work and trigger processing. Use for: waking idle agents, redirecting work, requesting new output.
+
+### Agent Lifecycle Management
+- **list_agents({})**: Returns all running agents. Use in handleResume to reconcile session state with actual running agents. Use in handleComplete to verify clean shutdown.
+- **Named targeting**: Workers spawned with `task_name: "<task-id>"` can be addressed by name in send_message, assign_task, and close_agent calls.
+
 ## Error Handling
 
 | Scenario | Resolution |

.codex/skills/team-ux-improve/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: team-ux-improve
 description: Unified team skill for UX improvement. Systematically discovers and fixes UI/UX interaction issues including unresponsive buttons, missing feedback, and state refresh problems. Uses team-worker agent architecture with roles/ for domain logic. Coordinator orchestrates pipeline, workers are team-worker agents. Triggers on "team ux improve".
-allowed-tools: spawn_agent(*), wait_agent(*), send_input(*), close_agent(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*), mcp__ace-tool__search_context(*), mcp__ccw-tools__read_file(*), mcp__ccw-tools__write_file(*), mcp__ccw-tools__edit_file(*), mcp__ccw-tools__team_msg(*)
+allowed-tools: spawn_agent(*), wait_agent(*), send_message(*), assign_task(*), close_agent(*), list_agents(*), report_agent_job_result(*), request_user_input(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*), mcp__ace-tool__search_context(*), mcp__ccw-tools__read_file(*), mcp__ccw-tools__write_file(*), mcp__ccw-tools__edit_file(*), mcp__ccw-tools__team_msg(*)
 ---
 
 # Team UX Improve
@@ -63,7 +63,8 @@ Before calling ANY tool, apply this check:
 
 | Tool Call | Verdict | Reason |
 |-----------|---------|--------|
-| `spawn_agent`, `wait_agent`, `close_agent`, `send_input` | ALLOWED | Orchestration |
+| `spawn_agent`, `wait_agent`, `close_agent`, `send_message`, `assign_task` | ALLOWED | Orchestration |
+| `list_agents` | ALLOWED | Agent health check |
 | `request_user_input` | ALLOWED | User interaction |
 | `mcp__ccw-tools__team_msg` | ALLOWED | Message bus |
 | `Read/Write` on `.workflow/.team/` files | ALLOWED | Session state |
@@ -94,6 +95,8 @@ Coordinator spawns workers using this template:
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: "<task-id>",
+  fork_context: false,
   items: [
     { type: "text", text: `## Role Assignment
 role: <role>
@@ -117,7 +120,21 @@ pipeline_phase: <pipeline-phase>` },
 })
 ```
 
-After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ id })` each worker.
+After spawning, use `wait_agent({ targets: [...], timeout_ms: 900000 })` to collect results, then `close_agent({ target })` each worker.
+
+
+### Model Selection Guide
+
+UX improvement has an explorer utility member and a scan-diagnose-design-implement-test pipeline.
+
+| Role | reasoning_effort | Rationale |
+|------|-------------------|-----------|
+| explorer | low | File reading and pattern summarization only (utility member) |
+| scanner | medium | Broad UI/UX issue scanning, pattern-based detection |
+| diagnoser | high | Root cause diagnosis of UX issues requires deep analysis |
+| designer | high | UX improvement design needs careful user experience reasoning |
+| implementer | high | UI code changes must preserve existing behavior |
+| tester | medium | Test execution follows defined UX verification criteria |
 
 ## User Commands
 
@@ -152,6 +169,50 @@ After spawning, use `wait_agent({ ids: [...], timeout_ms: 900000 })` to collect
     └── anti-patterns/
 ```
 
+## v4 Agent Coordination
+
+### Message Semantics
+
+| Intent | API | Example |
+|--------|-----|---------|
+| Queue supplementary info (don't interrupt) | `send_message` | Send explorer findings to running scanner |
+| Assign implementation from design | `assign_task` | Assign IMPL task from designer output |
+| Check running agents | `list_agents` | Verify agent health during resume |
+
+### Agent Health Check
+
+Use `list_agents({})` in handleResume and handleComplete:
+
+```
+// Reconcile session state with actual running agents
+const running = list_agents({})
+// Compare with meta.json active tasks
+// Reset orphaned tasks (in_progress but agent gone) to pending
+```
+
+### Named Agent Targeting
+
+Workers are spawned with `task_name: "<task-id>"` enabling direct addressing:
+- `send_message({ target: "SCAN-001", items: [...] })` -- send explorer findings to scanner
+- `assign_task({ target: "IMPL-001", items: [...] })` -- assign UX fix from designer output
+- `close_agent({ target: "TEST-001" })` -- cleanup after UX testing
+
+### Explorer-Assisted Scanning
+
+Coordinator spawns explorer (utility member) early to gather codebase context, then sends findings to scanner via upstream context:
+```
+// Explorer results inform scanner's scope
+send_message({
+  target: "SCAN-001",
+  items: [
+    { type: "text", text: `## Explorer Findings
+component_patterns: <explorer's component inventory>
+framework: <detected framework>
+ui_library: <detected UI library>` }
+  ]
+})
+```
+
 ## Error Handling
 
 | Scenario | Resolution |

.codex/skills/team-ux-improve/roles/coordinator/commands/monitor.md

@@ -79,6 +79,16 @@ Output status -- do NOT advance pipeline.
 
 ## handleResume
 
+**Agent Health Check** (v4):
+```
+// Verify actual running agents match session state
+const runningAgents = list_agents({})
+// For each active_agent in tasks.json:
+//   - If agent NOT in runningAgents -> agent crashed
+//   - Reset that task to pending, remove from active_agents
+// This prevents stale agent references from blocking the pipeline
+```
+
 1. Audit task list for inconsistencies:
    - Tasks stuck in "in_progress" -> reset to "pending"
    - Tasks with completed blockers but still "pending" -> include in spawn list
@@ -101,6 +111,7 @@ Find ready tasks, spawn workers, STOP.
 ```
 spawn_agent({
   agent_type: "team_worker",
+  task_name: taskId,  // e.g., "SCAN-001" — enables named targeting
   items: [{
     description: "Spawn <role> worker for <task-id>",
     team_name: "ux-improve",
@@ -133,9 +144,33 @@ Inner loop roles: implementer (inner_loop: true)
 Single-task roles: scanner, diagnoser, designer, tester (inner_loop: false)
 
 5. Add to active_workers, update session, output summary, STOP
+6. Use `wait_agent({ targets: [<spawned-task-names>], timeout_ms: 900000 })` to wait for callbacks. If `result.timed_out`, mark tasks as `timed_out` and close agents. Use `close_agent({ target: taskId })` with task_name for cleanup.
+
+**Cross-Agent Supplementary Context** (v4):
+
+When spawning workers in a later pipeline phase, send upstream results as supplementary context to already-running workers:
+
+```
+// Example: Send scan results to running diagnoser
+send_message({
+  target: "<running-agent-task-name>",
+  items: [{ type: "text", text: `## Supplementary Context\n${upstreamFindings}` }]
+})
+// Note: send_message queues info without interrupting the agent's current work
+```
+
+Use `send_message` (not `assign_task`) for supplementary info that enriches but doesn't redirect the agent's current task.
 
 ## handleComplete
 
+**Cleanup Verification** (v4):
+```
+// Verify all agents are properly closed
+const remaining = list_agents({})
+// If any team agents still running -> close_agent each
+// Ensures clean session shutdown
+```
+
 Pipeline done. Generate report and completion action.
 
 1. Verify all tasks (including any fix-verify iterations) have status "completed"

.codex/skills/team-ux-improve/roles/coordinator/role.md

@@ -6,7 +6,7 @@ UX Improvement Team coordinator. Orchestrate pipeline: analyze -> dispatch -> sp
 
 **You are a dispatcher, not a doer.** Your ONLY outputs are:
 - Session state files (`.workflow/.team/` directory)
-- `spawn_agent` / `wait_agent` / `close_agent` / `send_input` calls
+- `spawn_agent` / `wait_agent` / `close_agent` / `send_message` / `assign_task` calls
 - Status reports to the user / `request_user_input` prompts
 
 **FORBIDDEN** (even if the task seems trivial):
@@ -35,6 +35,8 @@ WRONG: Edit/Write on project source files            — worker work
 - Handle wisdom initialization and consolidation
 - Maintain session state persistence
 - **Always proceed through full Phase 1-5 workflow, never skip to direct execution**
+- Use `send_message` for supplementary context (non-interrupting) and `assign_task` for triggering new work
+- Use `list_agents` for session resume health checks and cleanup verification
 
 ### MUST NOT
 - Execute worker domain logic directly (scanning, diagnosing, designing, implementing, testing)
@@ -144,6 +146,16 @@ Delegate to `@commands/monitor.md#handleSpawnNext`:
    })
    ```
 
+## v4 Coordination Patterns
+
+### Message Semantics
+- **send_message**: Queue supplementary info to a running agent. Does NOT interrupt current processing. Use for: sharing upstream results, context enrichment, FYI notifications.
+- **assign_task**: Assign new work and trigger processing. Use for: waking idle agents, redirecting work, requesting new output.
+
+### Agent Lifecycle Management
+- **list_agents({})**: Returns all running agents. Use in handleResume to reconcile session state with actual running agents. Use in handleComplete to verify clean shutdown.
+- **Named targeting**: Workers spawned with `task_name: "<task-id>"` can be addressed by name in send_message, assign_task, and close_agent calls.
+
 ## Error Handling
 
 | Error | Resolution |