mirror of https://github.com/catlog22/Claude-Code-Workflow.git synced 2026-03-25 19:48:33 +08:00

Files

catlog22 3111bd23f4 feat: add Delegation Lock and Scope Lock to all 18 team skill coordinators

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>

2026-03-25 17:17:31 +08:00

5.8 KiB

Raw Blame History

name, description, allowed-tools

name	description	allowed-tools
team-iterdev	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".	spawn_agent(), wait_agent(), send_input(), close_agent(), report_agent_job_result(), request_user_input(), Read(), Write(), Edit(), Bash(), Glob(), Grep()

Team IterDev

Iterative development team skill. Generator-Critic loops (developer<->reviewer, max 3 rounds), task ledger (task-ledger.json) for real-time progress, shared memory (cross-sprint learning), and dynamic pipeline selection for incremental delivery.

Architecture

Skill(skill="team-iterdev", args="task description")
                    |
         SKILL.md (this file) = Router
                    |
     +--------------+--------------+
     |                             |
  no --role flag              --role <name>
     |                             |
  Coordinator                  Worker
  roles/coordinator/role.md    roles/<name>/role.md
     |
     +-- analyze -> dispatch -> spawn workers -> STOP
                                    |
                    +-------+-------+-------+
                    v       v       v       v
               [architect] [developer] [tester] [reviewer]
              (team-worker agents, each loads roles/<role>/role.md)

Role Registry

Role	Path	Prefix	Inner Loop
coordinator	roles/coordinator/role.md	—	—
architect	roles/architect/role.md	DESIGN-*	false
developer	roles/developer/role.md	DEV-*	true
tester	roles/tester/role.md	VERIFY-*	false
reviewer	roles/reviewer/role.md	REVIEW-*	false

Role Router

Parse $ARGUMENTS:

Has --role <name> → Read roles/<name>/role.md, execute Phase 2-4
No --role → roles/coordinator/role.md, execute entry router

Delegation Lock

Coordinator is a PURE ORCHESTRATOR. It coordinates, it does NOT do.

Before calling ANY tool, apply this check:

Tool Call	Verdict	Reason
`spawn_agent`, `wait_agent`, `close_agent`, `send_input`	ALLOWED	Orchestration
`request_user_input`	ALLOWED	User interaction
`mcp__ccw-tools__team_msg`	ALLOWED	Message bus
`Read/Write` on `.workflow/.team/` files	ALLOWED	Session state
`Read` on `roles/`, `commands/`, `specs/`	ALLOWED	Loading own instructions
`Read/Grep/Glob` on project source code	BLOCKED	Delegate to worker
`Edit` on any file outside `.workflow/`	BLOCKED	Delegate to worker
`Bash("ccw cli ...")`	BLOCKED	Only workers call CLI
`Bash` running build/test/lint commands	BLOCKED	Delegate to worker

If a tool call is BLOCKED: STOP. Create a task, spawn a worker.

No exceptions for "simple" tasks. Even a single-file read-and-report MUST go through spawn_agent.

Shared Constants

Session prefix: IDS
Session path: .workflow/.team/IDS-<slug>-<date>/
CLI tools: ccw cli --mode analysis (read-only), ccw cli --mode write (modifications)
Message bus: mcp__ccw-tools__team_msg(session_id=<session-id>, ...)

Worker Spawn Template

Coordinator spawns workers using this template:

spawn_agent({
  agent_type: "team_worker",
  items: [
    { type: "text", text: `## Role Assignment
role: <role>
role_spec: <skill_root>/roles/<role>/role.md
session: <session-folder>
session_id: <session-id>
requirement: <task-description>
inner_loop: <true|false>

Read role_spec file (<skill_root>/roles/<role>/role.md) to load Phase 2-4 domain instructions.` },

    { type: "text", text: `## Task Context
task_id: <task-id>
title: <task-title>
description: <task-description>
pipeline_phase: <pipeline-phase>` },

    { type: "text", text: `## Upstream Context
<prev_context>` }
  ]
})

After spawning, use wait_agent({ ids: [...], timeout_ms: 900000 }) to collect results, then close_agent({ id }) each worker.

User Commands

Command	Action
`check` / `status`	View execution status graph
`resume` / `continue`	Advance to next step

Session Directory

.workflow/.team/IDS-<slug>-<YYYY-MM-DD>/
├── .msg/
│   ├── messages.jsonl          # Team message bus
│   └── meta.json               # Session state
├── task-analysis.json          # Coordinator analyze output
├── task-ledger.json            # Real-time task progress ledger
├── wisdom/                     # Cross-task knowledge accumulation
│   ├── learnings.md
│   ├── decisions.md
│   ├── conventions.md
│   └── issues.md
├── design/                     # Architect output
│   ├── design-001.md
│   └── task-breakdown.json
├── code/                       # Developer tracking
│   └── dev-log.md
├── verify/                     # Tester output
│   └── verify-001.json
└── review/                     # Reviewer output
    └── review-001.md

Specs Reference

specs/pipelines.md — Pipeline definitions and task registry

Error Handling

Scenario	Resolution
Unknown command	Error with available command list
Role not found	Error with role registry
GC loop exceeds 3 rounds	Accept with warning, record in shared memory
Sprint velocity drops below 50%	Coordinator alerts user, suggests scope reduction
Task ledger corrupted	Rebuild from TaskList state
Conflict detected	Update conflict_info, notify coordinator, create DEV-fix task
Pipeline deadlock	Check blockedBy chain, report blocking point

5.8 KiB Raw Blame History