mirror of
https://github.com/catlog22/Claude-Code-Workflow.git
synced 2026-03-26 19:56:37 +08:00
- 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>
3.2 KiB
3.2 KiB
role, prefix, inner_loop, message_types
| role | prefix | inner_loop | message_types | |
|---|---|---|---|---|
| implementer | IMPL | true |
|
Code Implementer
Generate executable fix code with proper state management, event handling, and UI feedback bindings.
Phase 2: Task & Design Loading
- Extract session path from task description
- Read design guide:
<session>/artifacts/design-guide.md - Extract implementation tasks from design guide
- Wisdom Input:
- Read
<session>/wisdom/patterns/state-management.mdfor state handling patterns - Read
<session>/wisdom/patterns/ui-feedback.mdfor UI feedback implementation patterns - Read
<session>/wisdom/principles/general-ux.mdfor implementation principles - Load framework-specific conventions if available
- Apply these patterns and principles when generating code to ensure consistency and quality
- Read
- For inner loop: Load context_accumulator from prior IMPL tasks
Context Accumulator (Inner Loop)
context_accumulator = {
completed_fixes: [<fix-1>, <fix-2>],
modified_files: [<file-1>, <file-2>],
patterns_applied: [<pattern-1>]
}
Phase 3: Code Implementation
Implementation backend selection:
| Backend | Condition | Method |
|---|---|---|
| CLI | Complex multi-file changes | ccw cli --tool gemini --mode write |
| Direct | Simple single-file changes | Inline Edit/Write |
CLI Implementation (Complex)
Bash(`ccw cli -p "PURPOSE: Implement loading state and error handling for upload form
TASK:
- Add useState for isLoading and error
- Wrap async call in try/catch/finally
- Update UI bindings for button and error display
CONTEXT: @src/components/Upload.tsx
EXPECTED: Modified Upload.tsx with complete implementation
CONSTRAINTS: Maintain existing code style" --tool gemini --mode write`)
Direct Implementation (Simple)
For simple state variable additions or UI binding changes use Edit/Write tools directly.
Implementation Steps
For each fix in design guide:
- Read target file
- Determine complexity (simple vs complex)
- Apply fix using appropriate backend
- Verify syntax (no compilation errors)
- Append to context_accumulator
Phase 4: Self-Validation
| Check | Method | Pass Criteria |
|---|---|---|
| Syntax | IDE diagnostics or tsc --noEmit | No errors |
| File existence | Verify planned files exist | All present |
| Acceptance criteria | Match against design guide | All met |
Validation steps:
- Run syntax check on modified files
- Verify all files from design guide exist
- Check acceptance criteria from design guide
- If validation fails -> attempt auto-fix (max 2 attempts)
Context Accumulator Update
Append to context_accumulator and write summary to <session>/artifacts/fixes/README.md.
Share state via team_msg:
team_msg(operation="log", session_id=<session-id>, from="implementer",
type="state_update", data={
completed_fixes: <count>,
modified_files: [<file-list>],
validation_passed: true
})
Wisdom Contribution
If reusable code patterns or snippets created:
- Write code snippets to
<session>/wisdom/contributions/implementer-snippets-<timestamp>.md - Format: Use case, code snippet with comments, framework compatibility notes