update-agent-learnings — Claude Code skill: installation
The update-agent-learnings skill turns one confirmed lesson from a session into lasting agent guidance and syncs it to relevant agent files. Use it when a behavior rule should carry over to future sessions.
no license 160 installs ★ 5 updated 2026-08-31
npx skills add ulpi-io/skills --skill update-agent-learnings -a claude-codeWhat update-agent-learnings does
The skill identifies one actionable rule from the current session, assigns it a scope, and asks the user to approve the wording and files before making changes. Once approved, it updates the central learnings file and copies the regenerated ## Learnings section to the agent definitions it finds. The central file remains the single source of truth.
It is meant for persistent agent behavior guidance, such as testing practices or rules for a particular role. It does not handle one-off implementation notes, workflow rules for CLAUDE.md, or requests to rewrite an agent prompt immediately. Because the workflow changes durable prompts, it will not write anything without explicit user approval. Claude Code Only rules stay in the central file and are excluded from subagent prompts.
What it helps with
- Store one validated agent-behavior rule
- Classify a lesson as global or agent-specific
- Sync
## Learningsto the relevant agent files - Keep
Claude Code Onlyguidance out of subagent prompts
Example prompts for the agent
- Save the lesson from this session about when to run tests, and show me the wording first.
- Record this rule for the
revieweragent and sync the matching agent files. - Check whether this belongs in persistent agent guidance, then ask before editing anything.
How to install update-agent-learnings
1Skills CLI (Vercel)
npx skills add ulpi-io/skills --skill update-agent-learnings -a claude-code2Manually: put the files in place
~/.claude/skills/update-agent-learnings/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/update-agent-learnings/
1Skills CLI (Vercel)
npx skills add ulpi-io/skills --skill update-agent-learnings -a codex2Manually: put the files in place
.agents/skills/update-agent-learnings/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add ulpi-io/skills --skill update-agent-learnings -a cursor2Manually: put the files in place
.cursor/skills/update-agent-learnings/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install ulpi-io/skills update-agent-learnings2Manually: put the files in place
.github/skills/update-agent-learnings/SKILL.md1Skills CLI (Vercel)
npx skills add ulpi-io/skills --skill update-agent-learnings -a gemini-cli2Platform CLI
gemini skills install https://github.com/ulpi-io/skills --path update-agent-learningsManually: put the files in place
.devin/skills/update-agent-learnings/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add ulpi-io/skills --skill update-agent-learnings -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about update-agent-learnings
How do I install update-agent-learnings in Claude Code?
Run npx skills add ulpi-io/skills --skill update-agent-learnings -a claude-code. The skill file is installed at ~/.claude/skills/update-agent-learnings/SKILL.md.
What does update-agent-learnings do?
It extracts one agent-behavior rule from a session, confirms its wording and scope with the user, then updates the central learnings file and relevant agent definitions.
Does the skill edit agent files without approval?
No. It requires the user's explicit approval before changing the central learnings file or any agent file.