Agent Self-Improvement Log
Paste the following prompt into your AI chat to install this skill:
Please follow https://skillhub.cn/install/skillhub.md to install @user_be2168d6/11.
About this skill
Problem
AI coding assistants often hit session-specific pitfalls: failed commands, wrong package managers, API behavior that differs from documentation, or user corrections to mistaken assumptions. If that experience only lives in the current context, it disappears after the session ends, and the next task can repeat the same mistakes.
How it works
The skill turns experience into auditable Markdown records: errors go to .learnings/ERRORS.md, corrections, knowledge gaps, and best practices go to .learnings/LEARNINGS.md, and missing capabilities go to .learnings/FEATURE_REQUESTS.md. Entries can use a TYPE-YYYYMMDD-XXX ID format and track Status, priority, and See Also links. Learnings that apply across files or sessions can be distilled into CLAUDE.md, AGENTS.md, or .github/copilot-instructions.md; in OpenClaw workspaces, durable guidance can also go to SOUL.md or TOOLS.md. When the same issue recurs, Pattern-Key, Recurrence-Count, and a time window help decide whether to promote it into a concise system-level rule rather than a long incident note.
Boundaries
It is not an automatic fixer and does not replace tests, CI, or code review. Record quality depends on including reproduction steps, related files, and concrete fix suggestions. Teams need to decide whether .learnings/ stays local or is tracked in Git; for single-project use, keep permanent rules in a small set of project memory files instead of turning one-off errors into lasting conventions.
Use Cases
- Debugging CI failures in Claude Code, log pnpm dependency errors to .learnings/ERRORS.md to avoid re-guessing the package manager.
- After repeated user corrections on API fields, capture knowledge gaps and best practices, then distill them into AGENTS.md conventions.
- In OpenClaw multi-session work, record Git auth and tool gotchas in TOOLS.md so later sessions see stable guidance.
- For recurring simplification patterns, track Pattern-Key counts and promote concise rules into CLAUDE.md after the threshold.
Best For
- Claude Code users maintaining backend repos who want repeated dependency and API pitfalls written into CLAUDE.md.
- OpenClaw multi-session engineers who need tool gotchas and workflow rules synced into TOOLS.md and AGENTS.md.
- Platform engineers owning Copilot rules who want shared team conventions captured in .github/copilot-instructions.md.
- Independent developers debugging agent behavior who want to separate one-off fixes from durable, promotable experience.
Related Skills
A free matrix publishing assistant for Skill developers, guiding distribution to Tencent SkillHub, XiaLiao, XiaYou SkillHub, GitHub, and ClawHub with create, update, and delete safeguards.
A guide for writing Seedance 2.0 video prompts, covering @ references, camera moves, time segments, audio direction, and common pitfalls.
Collects daily conversations, task logs, and errors, then runs a nine-module review to surface blind spots, generate actions, and update memory.
Local long-term memory for AI agents that appends key facts and daily logs, supports cross-session context and search recall, and auto-archives old diaries to gzip files.