capabilityevolver1037
от opencloseopenclose
Web & Frontend DevelopmentКратко
A self-evolution engine for AI agents.
Описание
Что делает этот навык
A meta-skill that lets OpenClaw agents inspect their own runtime history, detect failures, and autonomously write patches or generate new skills. Includes a continuous self-healing loop mode and a structured knowledge accumulation system called Ascension Protocol.
Instead of manually reviewing logs and patching agent code, the evolver closes the feedback loop autonomously without requiring a separate monitoring stack.
Когда его использовать
- Automatically patching a crashed agent after a failed run
- Extracting recurring error patterns into a persistent knowledge base
- Generating a new skill file to handle a gap discovered at runtime
- Running a continuous self-healing loop overnight on an unstable agent
- Auditing agent history to surface inefficiencies before a production deploy
Пример сценария
User runs: node skills/capability-evolver/index.js
- 1
Scans memory and history files for errors, crashes, and user corrections
- 2
Classifies each finding as fixable code, a learnable rule, or a critical doc update
- 3
Writes patches to broken skill files or generates new skill stubs in skills/
- 4
Crystallizes extracted lessons into memory/KNOWLEDGE_BASE/LESSONS_LEARNED.md
- 5
Commits all changes to Git via workspace sync
Updated skill files, a populated LESSONS_LEARNED.md, and a Git commit capturing all autonomous changes
Установка
clawhub install opencloseopenclose/capabilityevolver1037Выполните команду CLI или вставьте промпт в Claude Code, Cursor, Codex — любой кодинг-агент, которым пользуетесь.