| name | agentic:skill:code |
| description | Use when implementing code according to a technical plan. After planning, during development. |
Code Implementation
Implement strictly according to the Architect's technical plan.
Source of Truth
technical-context.md
technical-plan.md
Evidence Policy (HARD)
You MUST provide evidence of verification:
- List exact commands you ran (lint/typecheck/tests)
- Summarize results
- If cannot run, state precisely why and what CI will run
- Never claim "tests pass" without evidence
Implementation Discipline
- Keep changes PR-sized (aligned with
TASK-*)
- Do not introduce new requirements
- Respect architectural boundaries and dependency direction
- Prefer small files, clear naming, single responsibility
- Update/add tests per Verification Matrix
- Never disable checks to "make it pass"
Output Structure
When responding:
| Section | Content |
|---|
| Summary | What changed |
| Plan | Which tasks implemented |
| Changes | By file |
| Tests | Added/updated |
| Commands run + results | Evidence |
| Compliance check | Map to AC-* IDs |
| Risks & follow-ups | |
Implementation Log (Optional but Recommended)
.claude/_agentic_output/task/{epicNumber}-EPIC-{epicName}/US-{usName}-{usNumber}/implementation-log.md
Contents:
- Commits/SHAs (if available)
- Commands run + outputs summary
- Deviations from plan (if any) + rationale
GitHub Integration
After implementation complete:
- Run
/create-pr with implementation-log.md as description
- PR auto-links to related user story issue