auto-coder

Autonomous agent: one trigger completes read spec → find task → code → test → persist progress.

Trigger

| User Says | Behavior | |-----------|----------| | "auto code" / "自动开发" | Next task, full cycle | | "auto code B2" | Specific task | | "auto code --no-commit" | Skip git commit |

---

Pipeline

Sync Spec → Find Task → Implement → Test (≤3 fix rounds) → Persist

Only pause at the very end for commit confirmation. Everything else runs autonomously.

⚠️ CRITICAL: ALL Python commands MUST run inside the project venv. Before executing ANY python or pytest command, activate the venv first:

.\.venv\Scripts\Activate.ps1

Verify by checking Get-Command python points to .venv\Scripts\python.exe. Never use system Python. Never skip this step.

---

1. Sync Spec

Activate venv first, then sync:

.\.venv\Scripts\Activate.ps1
python .github/skills/auto-coder/scripts/sync_spec.py

Then read the schedule file to get task statuses:

• Read .github/skills/auto-coder/specs/06-schedule.md

Task markers:

| Marker | Status | |--------|--------| | [ ] / ⬜ | Not started | | [~] / 🔶 / (进行中) | In progress | | [x] / ✅ / (已完成) | Completed |

---

2. Find Task

Priority: first IN_PROGRESS, then first NOT_STARTED. If user specified a task ID, use that directly.

Quick-check predecessor artifacts exist (file-level only). On mismatch, log warning and continue — only stop if the target task itself is blocked.

---

3. Implement

1. Read relevant spec from .github/skills/auto-coder/specs/:

   • Architecture: 05-architecture.md
   • Tech details: 03-tech-stack.md
   • Testing conventions: 04-testing.md

2. Extract from spec: inputs/outputs, design principles (Pluggable? Config-driven? Factory?), file list, acceptance criteria.

3. Plan files to create/modify before writing any code.

4. Code — mandatory standards:

   • Type hints on all signatures
   • Google-style docstrings on public APIs
   • No hardcoded values (use config)
   • Single responsibility, short functions
   • Error handling for external integrations

5. Write tests alongside code:

   • tests/unit/test_<module>.py or tests/integration/ per spec
   • Naming: test_<func>_<scenario>_<expected>
   • Mock external deps in unit tests

6. Self-review before running tests: all planned files exist, type hints present, no hardcoded values, tests import correctly.

---

4. Test & Auto-Fix

Round 0..2:
  Run pytest on relevant test file
  If pass → go to step 5
  If fail → analyze error, apply fix, re-run

Round 3 still failing → STOP, show failure report to user

---

5. Persist

1. Update DEV_SPEC.md (global file): change task marker [ ] → [x]
2. Re-sync: python .github/skills/auto-coder/scripts/sync_spec.py --force
3. Show summary & ask:

✅ [A3] 配置加载与校验 — done
   Files: src/core/settings.py, tests/unit/test_settings.py
   Tests: 8/8 passed
   Commit: feat(config): [A3] implement config loader

   "commit" → git add + commit
   "skip"   → end
   "next"   → commit + start next task

On "next", loop back to step 1 for the next task.

---

Guardrails

• One task per cycle, atomic commits
• Spec is single source of truth
• 3-round test fix limit
• Match existing codebase style
• MUST activate .venv before ANY python/pytest command — no exceptions. If unsure whether venv is active, run .\.venv\Scripts\Activate.ps1 again (idempotent)

---

Directory Structure

auto-coder/
├── SKILL.md              ← this file
├── .spec_hash            ← auto-generated hash
├── scripts/
│   └── sync_spec.py      ← splits DEV_SPEC.md into chapters
└── specs/                ← auto-generated chapter files
    ├── 01-overview.md
    ├── 02-features.md
    ├── 03-tech-stack.md
    ├── 04-testing.md
    ├── 05-architecture.md
    ├── 06-schedule.md
    └── 07-future.md

All paths are self-contained. This skill has no external dependencies on other skills.