メニュー
This skill should be used when the user says "learn from that mistake", "remember this for next time", "add that to CLAUDE.md", "don't do that again", or wants to codify a correction or lesson into CLAUDE.md so it persists across sessions.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Use when writing documentation, READMEs, explanations, blog posts, or any prose meant for humans to read
This skill should be used when the user asks to "rebase all open PRs", "update all PRs with latest main", "make sure every open PR is up to date with main", "sync all open PRs onto main", "rebase all my PRs", or wants to bring every open pull request current with its base branch by rebasing in parallel git worktrees, resolving conflicts with a team of agents, and force-pushing.
Groom and prioritize a project backlog — verify each candidate issue's claims against the current codebase, pick the most important N, rewrite them to a ready-to-work standard (Problem/Goals/Decisions/Outcomes/Verification), move them to Ready on the board, and close obsolete issues. Use when asked to "groom the backlog", "triage the backlog", "prioritize issues", "move issues to ready", or "what should we work on next".
This skill should be used when the user asks to "summarize a note", "summarize this article", "replace #summarize-later", "add a summary to a note", "generate takeaways", or mentions summarizing Obsidian notes. Replaces the #summarize-later tag with a high-level summary and actionable takeaways.
This skill should be used when the user asks to "update the changelog", "add changelog entries", "what changed since last release", or when updating CHANGELOG.md with user-facing changes. Follows Keep a Changelog format and filters out internal-only commits.
This skill should be used when the user asks to "improve a prompt", "make this prompt better", "optimize this prompt", "refine this system prompt", or wants to apply concept elevation to compress and clarify LLM instructions.
| name | learn |
| user-invokable | true |
| description | This skill should be used when the user says "learn from that mistake", "remember this for next time", "add that to CLAUDE.md", "don't do that again", or wants to codify a correction or lesson into CLAUDE.md so it persists across sessions. |
Review the conversation to identify the most recent mistake or correction, then codify it as a rule in the appropriate CLAUDE.md.
Identify the mistake: Scan the conversation history for the most recent correction, mistake, or suboptimal behavior. Summarize what went wrong in one sentence.
Draft a rule: Write a concise, actionable rule that prevents this mistake in the future. The rule should be:
Choose the right file: Determine which CLAUDE.md to update:
./CLAUDE.md): for lessons specific to this project's codebase, conventions, or tooling~/.claude/CLAUDE.md): for lessons that apply across all projects (general coding habits, communication style, workflow)Check for duplicates: Read the target CLAUDE.md and verify the rule doesn't already exist or overlap with an existing rule. If a related rule exists, refine it rather than adding a duplicate.
Add the rule: Append the rule to the most appropriate existing section, or create a new ## Learned Rules section if no section fits. Keep the file well-organized.
Show the user: Display the exact rule added and which file was updated.