Unified tool replacing manual schema reading (cat schema) across all agents
and skills. Supports 5 commands: init (skeleton), set (incremental field
setting with instant validation), validate (full structural + semantic),
merge (dedup multiple JSONs), info (compact schema summary).
Registers 24 schemas in schema-registry.ts. Updates all agent/skill/command
files to use json_builder info/init/validate instead of cat schema references.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The template generated unquoted description values containing `: ` sequences,
causing "mapping values are not allowed in this context" YAML errors in
generated SKILL.md files.
Closes#143
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add multi_agent_v2=true to Codex config documentation in both
README.md and README_CN.md, required for v4 agent API features
(fork_context, task_name, send_message, assign_task, list_agents).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
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>
When accessing the dashboard remotely (--host 0.0.0.0), all API calls
returned 401 because the frontend never acquired an auth_token cookie.
Added initializeAuth() that fetches /api/auth/token before React renders,
so the browser has the cookie before any TanStack Query hooks fire.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Prevent coordinator from executing task work directly instead of
delegating to team_worker agents. Three-layer enforcement:
- SKILL.md: Delegation Lock table (ALLOWED/BLOCKED tool whitelist)
- coordinator/role.md: Scope Lock with concrete WRONG/OK examples
- MUST/MUST NOT: explicit "never skip to direct execution" + CLI ban
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Two issues fixed:
1. Default perspective selection now auto-recommends Technical + Architectural
when dimensions >= 2, instead of always defaulting to single comprehensive
view (which caused Phase B to never trigger).
2. Phase B pseudocode replaced forEach sequential pattern with explicit
parallel Agent() call pattern — all perspective agents must be launched
in the same response block for concurrent execution.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Mirror the same Findings Coverage Matrix mechanism added to the Claude
version. Insert Step 4.1 between Intent Coverage (4.0) and Consolidate
Insights (now 4.2) to ensure all actionable findings are mapped to a
disposition before generating recommendations.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add Phase 4 Step 2 "Findings Coverage Matrix" to ensure all actionable
findings from analysis rounds are mapped to a disposition (recommendation,
absorbed, deferred, or informational) before consolidation. This prevents
the issue where rich analysis discovers many actionable points but only
a subset makes it into final recommendations.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Batch now has a gear toggle for Claude advanced fields:
- Quick settings: color, permission, memory, maxTurns, background, isolation
- Tools row: tools, disallowedTools, skills
- When target=All, note indicates claude-only fields apply to claude agents only
- When target=Codex, advanced section hidden entirely
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Batch controls now distinguish codex vs claude agents:
- Type selector (All/Codex/Claude) filters targets and shows count
- Effort options adapt per type (no 'max' for codex-only)
- Color and Permission batch fields shown when claude agents targeted
- Claude-only fields applied individually per agent via PUT endpoint
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Support all Claude agent frontmatter fields (color, permissionMode, memory,
maxTurns, background, isolation, tools, disallowedTools, skills, mcpServers,
hooks) with MCP server picker, hooks editor, and progressive disclosure UI.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Convert 20 agent .md files to Codex-native .toml format with proper
metadata (name, description, model, sandbox_mode, developer_instructions).
Update all 19 skill files to use agent_type references instead of .md
file paths. Remove "Read role definition" bootstrapping step from spawn
messages since TOML developer_instructions replaces it.
Agent format: YAML frontmatter + body → TOML with inline instructions
Calling schema: agent: "path.md" → agent_type: "toml_name"
Sandbox: read-only for exploration agents, workspace-write for executors
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace nested settings structure with flat endpoints array for
export/import. Add conflict strategy options (skip/overwrite/merge)
and skipInvalid/disableImported flags.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Extract generalized tasks.json schema from team-lifecycle-v4 to shared
location (~/.ccw/workflows/cli-templates/schemas/). Codex needs explicit
schema since it uses file-based task state (unlike Claude Code's built-in
TaskCreate/TaskUpdate API). Simplify lifecycle-v4 tasks-schema.md to
reference the shared schema and only document skill-specific extensions.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Delete 21 old team skill directories using CSV-wave pipeline pattern (~100+ files)
- Delete old team-lifecycle (v3) and team-planex-v2
- Create generic team-worker.toml and team-supervisor.toml (replacing tlv4-specific TOMLs)
- Convert 19 team skills from Claude Code format (Agent/SendMessage/TaskCreate)
to Codex format (spawn_agent/wait_agent/tasks.json/request_user_input)
- Update team-lifecycle-v4 to use generic agent types (team_worker/team_supervisor)
- Convert all coordinator role files: dispatch.md, monitor.md, role.md
- Convert all worker role files: remove run_in_background, fix Bash syntax
- Convert all specs/pipelines.md references
- Final state: 20 team skills, 217 .md files, zero Claude Code API residuals
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Phase 3 Record-Before-Continue now checks for unresolved alternatives
in Proposed solutions and surfaces them to user. Phase 4 Consolidation
adds Solution Readiness Gate to flag recommendations with ambiguity.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- lite-plan LP-Phase 4: split single "Review" into two selections (Code Review + Convergence Review)
- lite-execute: add Step 4 Code Review (agent/codex/gemini) with code-review.md artifact, Step 5 passes convergenceReviewTool
- lite-test-review: rename reviewTool → convergenceReviewTool, TR-Phase 2 focused on convergence criteria verification
- All autoYes paths default both reviews to Skip
- Data structures updated across all three files for consistency
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
