--- name: auto-error-fixer description: | Autonomous error detection and self-healing code evolution pattern. Monitors a log file continuously, detects fixable Python errors, deduplicates by signature hash, creates dev-tasks.json, and launches plan-to-code-async for autonomous fix execution with real-time Telegram progress updates. Use when adding self-healing capability to a Python service, bot, or daemon. Reusable pattern — works in any project that has a log file and uses plan-to-code-async. allowed-tools: Read, Write, Bash, Glob, Grep --- # Auto Error Fixer — Self-Healing Code Evolution Adds autonomous error detection and fixing to any Python service. When a fixable error appears in the log, the system creates a dev-tasks.json and launches `plan-to-code-async` — which fixes the code and sends Telegram updates automatically. ## Quick Start ```bash # 1. Copy the auto_evolver.py template to your project cp ~/.claude/skills/auto-error-fixer/references/auto_evolver_template.py \ your_project/services/darwin/auto_evolver.py # 2. Initialize in your orchestrator/main loop from services.darwin.auto_evolver import get_auto_evolver auto_evolver = get_auto_evolver() auto_evolver.initialize() # seeks log to current end (ignores old errors) # 3. Call scan() every ~90s from your background loop await auto_evolver.scan() # 4. That's it — errors trigger plan-to-code-async automatically ``` ## How It Works ``` Log file (new lines every 90s) │ ▼ _extract_errors() ← groups ERROR + Traceback blocks │ ▼ _is_fixable() ← skip network/SSL/transient, keep code bugs │ ▼ _signature() + cooldown check ← deduplicate, prevent fix loops │ ▼ Write dev-tasks.json ← state/darwin/fixes// │ ▼ subprocess.Popen(launch.sh) ← detached, start_new_session=True │ ▼ plan-to-code-async ← fixes code + sends Telegram progress updates ``` ## Error Classification **Skip (transient / external — not fixable):** - `TelegramRetryAfter`, `Network is unreachable`, `Connection refused` - `httpx`, `aiohttp`, `SSL`, `TimeoutError`, `Rate limit`, `Flood control` **Fix (code bugs — autonomous repair):** - `AttributeError`, `NameError`, `KeyError`, `TypeError`, `ValueError` - `ImportError`, `SyntaxError`, `RuntimeError`, `UnboundLocalError` - `IndexError`, `AssertionError`, `exit code -9`, `SIGKILL`, `NoneType object` ## Configuration Constants ```python ERROR_COOLDOWN_S = 600 # 10 min between retries for same error signature MAX_FIXES_PER_DAY = 5 # safety cap on autonomous fixes SCAN_INTERVAL = 90 # seconds between log scans (set in caller loop) ``` ## Integrating into an Orchestrator Loop ```python # In your background _loop(): _auto_evolve_interval = 90 _last_auto_evolve_check = 0.0 while self._running: now = time.time() if self._auto_evolver and now - _last_auto_evolve_check >= _auto_evolve_interval: _last_auto_evolve_check = now try: await self._auto_evolver.scan() except Exception as e: logger.warning(f"auto_evolver scan failed: {e}") await asyncio.sleep(30) ``` ## dev-tasks.json Format Written Per Fix ```json { "project": "AutoEvolver Fix: AttributeError (20260222_143012)", "project_root": "/absolute/path/to/project", "test_framework": "pytest", "instructions": "Fix production bug. Read files before editing. Never break existing functionality.", "tasks": [{ "id": "1-fix-attributeerror", "step": 1, "title": "Fix AttributeError in production", "description": "Error log + traceback + file hints + restart instructions", "files_to_modify": ["telegram_bot/bot.py"], "status": "pending", "attempts": 0, "errors": [] }] } ``` ## Restart After Fix The task description instructs Claude to restart the bot if core files were modified: ``` If bot.py or any core service file was modified, restart: /path/to/scripts/restart-bot.sh 'Auto-Evolver: ' ``` ## State Files | File | Purpose | |------|---------| | `state/darwin/fixes//dev-tasks.json` | Per-fix task file | | `state/darwin/auto_fixes.jsonl` | Audit log of all fix attempts | | `state/darwin/fix_cooldowns.json` | Persisted error cooldowns | ## Reference Implementation - Full source: See [references/auto_evolver_template.py](references/auto_evolver_template.py) - Production instance: `~/teleclaudious/telegram_bot/services/darwin/auto_evolver.py` - Orchestrator integration: `~/teleclaudious/telegram_bot/services/darwin/orchestrator.py`