Claude-skill-registry knowledge-sync
install
source · Clone the upstream repo
git clone https://github.com/majiayu000/claude-skill-registry
Claude Code · Install into ~/.claude/skills/
T=$(mktemp -d) && git clone --depth=1 https://github.com/majiayu000/claude-skill-registry "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/data/knowledge-sync" ~/.claude/skills/majiayu000-claude-skill-registry-knowledge-sync && rm -rf "$T"
manifest:
skills/data/knowledge-sync/SKILL.mdsource content
Knowledge Sync
Keep skills and documentation in sync with codebase changes.
Announce: "I'm using knowledge-sync to update team knowledge after these changes."
When to Use
- After completing an OpenSpec change (especially refactors)
- When you notice a skill references outdated code
- When a new pattern is established that should be documented
- When asked to run
/sync-knowledge
Process
Phase 1: Identify What Changed
Check recent changes:
git log --oneline -20 git diff HEAD~10 --name-only
Categorize changes by domain:
- Database:
changessupabase/db/ - Frontend:
changessrc/ - Edge Functions:
changessupabase/functions/ - Config:
, AGENTS.md, skill changesopencode.jsonc
Phase 2: Map Changes to Skills
Use the skill-pattern map to identify affected skills:
| Code Area | Affected Skills |
|---|---|
| , , , |
| , |
| |
| , |
| |
| |
Phase 3: Analyze Impact
For each affected skill, check:
- Do code examples still work? Read the actual code, compare to skill examples
- Are anti-patterns still relevant? Maybe they're now the correct pattern
- Are new patterns worth documenting? Repeated patterns should be captured
- Is the description still accurate? Trigger words should match current functionality
Phase 4: Ask Clarifying Questions
If something is confusing, ASK before assuming:
I noticed [observation]. Questions: 1. [Specific question about intent] 2. [Question about whether this is a new pattern] 3. [Question about whether old pattern should be deprecated] Please clarify before I update the skills.
Phase 5: Propose Updates
Present changes in structured format:
## Skill Updates Required ### [skill-name] **Reason:** [Why this needs updating] **Current content (outdated):**
[snippet]
**Proposed content:**
[snippet]
--- Approve these changes?
Phase 6: Apply Updates (with approval)
After approval:
- Update the affected SKILL.md files
- Update references/ if needed
- Verify skill still loads correctly
- Summarize what was changed
What NOT to Update
- Workflow skills (openspec-*, executing-tasks, etc.) - These are process, not code
- Glossary entries - Unless domain terms changed meaning
- External API references - Unless we changed how we use them
Creating New Skills
If a new pattern emerges that doesn't fit existing skills:
- Identify the pattern clearly
- Check if it should extend an existing skill or be new
- If new, propose the skill structure:
skill-name/ ├── SKILL.md └── references/ └── examples.md - Get approval before creating
Red Flags
- Don't update based on assumptions - Read the actual code
- Don't remove patterns without asking - They may still be valid
- Don't add every code change - Only patterns worth repeating
- Don't update skills during a refactor - Wait until it's complete
References
See
references/skill-pattern-map.md for the full mapping.