AI Agent Hub
Back to skills
Agent Self-Improvement Learning Log icon

Agent Self-Improvement Learning Log

AI Agent Updated 2026.08.30

Paste the following prompt into your AI chat to install this skill:

Please follow https://skillhub.cn/install/skillhub.md to install @user_bf4124be/1231231231231212 into your AI assistant.

About this skill

Problem

Coding agents can repeatedly hit the same failures: broken commands, stale API assumptions, or user corrections that disappear with the session. This skill turns ephemeral context into inspectable project memory.

How It Works

It initializes .learnings/ at the project root and routes entries into LEARNINGS.md, ERRORS.md, and FEATURE_REQUESTS.md. Each entry can carry status, priority, area tags, and reproduction details. Similar entries are linked with See Also, and recurring issues can be promoted into CLAUDE.md, AGENTS.md, or .github/copilot-instructions.md; OpenClaw workspaces may also write to SOUL.md and TOOLS.md. Hook reminders are optional.

Boundaries

It is for logging errors, corrections, feature requests, and stable patterns—not for storing secrets, tokens, or full source files. Cross-session sharing should only happen in trusted environments and should use sanitized summaries.

Use Cases

  • Record Bash failures with repro steps in `.learnings/ERRORS.md`
  • Log user API corrections as `correction` entries in `LEARNINGS.md`
  • Promote recurring pnpm/npm setup issues into `AGENTS.md`
  • Share sanitized learning summaries across trusted OpenClaw sessions

Best For

  • Engineers maintaining Claude Code projects who need corrections to survive sessions
  • OpenClaw agent engineers who want durable rules in `SOUL.md` or `TOOLS.md`
  • Platform engineers maintaining `AGENTS.md` who need recurring failures turned into workflow rules
  • Repo owners integrating Copilot who want shared agent context conventions