Log learnings and errors to markdown files for continuous improvement. Coding agents can later process these into fixes, and important learnings get promoted to project memory.
| Command/operation fails | Log to .learnings/ERRORS.md | | User corrects you | Log to .learnings/LEARNINGS.md with category correction | | User wants missing feature | Log to .learnings/FEATUREREQUESTS.md | | API/external tool fails | Log to .learnings/ERRORS.md with integration details |
| Knowledge was outdated | Log to .learnings/LEARNINGS.md with category knowledgegap | | Found better approach | Log to .learnings/LEARNINGS.md with category bestpractice | | Similar to existing entry | Link with See Also, consider priority bump | | Broadly applicable learning | Promote to CLAUDE.md, AGENTS.md, and/or .github/copilot-instructions.md |
Capta aprendizajes, errores y correcciones para permitir la mejora continua. Úselo cuando: (1) Un comando u operación falla inesperadamente, (2) El usuario corrige a Claude ("No, eso está mal...", "En realidad..."), (3) El usuario solicita una capacidad que no existe, (4) Una API o herramienta externa falla, (5) Claude se da cuenta de que su conocimiento está desactualizado o es incorrecto, (6) Se descubre un mejor enfoque para una tarea recurrente. También revise los aprendizajes antes de realizar tareas importantes. Fuente: hanzoskill/self-improving-agent.