refresh-claude
Skill16 starsChanged 10 months agoMITView on GitHub
Review recent changes and update CLAUDE.md to keep it relevant for future sessions. Use after significant changes or periodically to maintain context.
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 7 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 45
Read the existing CLAUDE.md and identify:
- Distinctive: line 53
Check for patterns not reflected in CLAUDE.md:
- Distinctive: line 130
Don't repeat what's in:
- Distinctive: line 136
Remove information that's:
- Distinctive: line 164
After adding major features
- Distinctive: line 166
After adding new tools or dependencies
- Distinctive: line 168
Before onboarding new team members
≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Refresh CLAUDE.md
You are refreshing the project's CLAUDE.md file to ensure it stays relevant and useful for future Claude Code sessions.
Purpose
CLAUDE.md serves as persistent context between sessions. When a project evolves, CLAUDE.md should evolve too. This skill:
- Reviews recent changes to the codebase
- Identifies new patterns, conventions, or important context
- Updates CLAUDE.md to reflect current state
- Removes outdated information
Process
Step 1: Analyze Recent Changes
# Recent commits
git log --oneline -20
# Changed files
git diff --stat HEAD~10..HEAD
# New files
git diff --name-status HEAD~10..HEAD | grep "^A"Look for:
- New features or modules
- New dependencies or tools
- Changed conventions or patterns
- New commands or scripts
- Architecture changes
Step 2: Review Current CLAUDE.md
Read the existing CLAUDE.md and identify:
- Outdated information
- Missing new features/tools
- Incorrect commands or paths
- Stale conventions
Step 3: Scan the Codebase
Check for patterns not reflected in CLAUDE.md:
# Check for new test patterns
ls -la **/test* **/*test* **/*spec* 2>/dev/null | head -20
# Check for new config files
ls -la *.config.* .* 2>/dev/null | head -20
# Check package.json/pyproject.toml/etc for new scripts
cat package.json 2>/dev/null | jq '.scripts' || trueStep 4: Update CLAUDE.md
Make targeted updates:
DO update:
- New commands (dev, test, build, lint)
- New project structure (directories, key files)
- New conventions discovered in code
- Important architectural decisions
- New dependencies that affect how to work with the code
DON'T add:
- Implementation details that change frequently
- Obvious things (e.g., "this is a JavaScript project" for a .js repo)
- Temporary workarounds
- Personal preferences not enforced by tooling
Step 5: Verify Changes
After updating, verify:
- Commands still work
- Paths are correct
- No duplicate sections
- Concise and scannable
Output Format
## CLAUDE.md Refresh Summary
### Changes Made
- Added: [what was added]
- Updated: [what was updated]
- Removed: [what was removed]
### Context Discovered
- [New pattern or convention found]
- [New tool or command found]
### Recommendations
- [Suggestions for project improvements]Guidelines
Keep It Concise
CLAUDE.md should be scannable. Every line should earn its place.
❌ "This project uses React, which is a JavaScript library for building user interfaces..."
✓ "React 18 with TypeScript. See src/components/ for patterns."Focus on What's Actionable
Include information Claude needs to work effectively:
✓ Commands to run
✓ Key file locations
✓ Conventions to follow
✓ Things to avoidAvoid Duplication
Don't repeat what's in:
- README.md (user-facing docs)
- Code comments (implementation details)
- Config files (tooling handles it)
Keep It Fresh
Remove information that's:
- No longer accurate
- Covered by tooling (linters, formatters)
- Too detailed for context
Example Updates
Adding a New Command
## Commands
+ npm run e2e # Run end-to-end testsAdding a Convention
## Conventions
+ API routes follow REST conventions in `src/api/`
+ Use `zod` for runtime validation at API boundariesRemoving Outdated Info
- ## Legacy
- The old auth system in `src/auth-old/` is deprecated.When to Run
- After adding major features
- After significant refactoring
- After adding new tools or dependencies
- Periodically (weekly/monthly) on active projects
- Before onboarding new team members
In this repository12 other files it gives its agents.
- docs/agents.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- changelog-writerSkill
- code-reviewSkill
- explain-codeSkill
- generate-testsSkill
- install-precommitSkill
- refactor-codeSkill
- release-checklistSkill
- review-mrSkill
- risk-registerSkill
- wiggumSkill
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

