- Home
- Skills
- Petekp
- Agent Skills
- Capture Learning
capture-learning_skill
- Makefile
2
GitHub Stars
1
Bundled Files
2 months ago
Catalog Refreshed
4 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill petekp/agent-skills --skill capture-learning- SKILL.md10.4 KB
Overview
This skill analyzes recent conversation context and captures concise, actionable learnings into either project-specific knowledge files or reusable skills/commands for cross-project use. It helps convert debugging outcomes, setup fixes, and discovered patterns into well-structured documentation entries and proposes changes for user confirmation before writing.
How this skill works
The skill reviews the recent chat to identify problem, root cause, solution, and prevention steps, then determines whether the insight is project-specific or general. For project-specific items it detects .claude/ CLAUDE.md or other docs and drafts a formatted entry; for general patterns it prepares updates to skills/commands. It always presents the target file, exact content, and diff preview and waits for explicit approval before applying changes.
When to use it
- After resolving a build, setup, or environment issue and you want to document the fix
- When you discover a non-obvious debugging step or diagnostic technique worth reusing
- If you identify a project-specific convention, dependency requirement, or CI quirk
- When a cross-project pattern or framework gotcha should become a reusable skill
- Whenever you tell the agent to “capture this learning”, “update the docs”, or “remember this for next time”
Best practices
- Extract concrete, actionable items: Problem → Cause → Solution (and Prevention) rather than vague summaries
- Decide scope first: project-specific entries go in .claude/ files; general patterns become skills/commands
- Prefer CLAUDE.md for general project notes, use troubleshooting.md or setup.md for targeted fixes
- Show the proposed file, section, and diff and require explicit user confirmation before writing
- Avoid recording secrets or sensitive data—document patterns without credentials
Example use cases
- Documenting that Node 20 requires
npm install --legacy-peer-depsfor this repo and adding it to .claude/CLAUDE.md - Capturing a multi-step debugging process (symptoms, diagnostics, root cause) into .claude/docs/troubleshooting.md
- Promoting a framework gotcha (e.g., React Server Components constraints) into a reusable cross-project skill
- Adding required environment variable instructions to .claude/docs/setup.md after a setup session
- Noting an internal code convention (ApiError usage) into CLAUDE.md under Patterns for future contributors
FAQ
List each learning separately, ask which to capture (or capture all), then process each with a confirmation step.
How are target files chosen?
Priority: .claude/CLAUDE.md (or root CLAUDE.md), then .claude/docs/*.md; create new files only if the user requests it.