| name | auto-sync |
| description | Keep a graphify knowledge graph current automatically: --watch mode (live rebuild on file save) or a post-commit git hook (rebuild per commit). TRIGGERS - graphify watch, auto-sync graph, keep graph updated, graphify git hook, graph out of date. |
| allowed-tools | Read, Bash, Glob, AskUserQuestion |
Auto-Sync the Graph
Self-Evolving Skill: This skill improves through use. If the hook behavior or --watch semantics drift — fix this file immediately, don't defer. Only update for real, reproducible issues.
Two mechanisms keep graphify-out/ in step with a changing corpus. They don't conflict — pick per repo temperament.
Prerequisite: engine installed (graphify-tools:setup) and an initial graph built (graphify-tools:build-graph).
Option A — post-commit git hook (recommended default)
cd <repo> && graphify hook install
- Rebuilds the graph once per commit; no background process
- Safe alongside existing hooks (appends, doesn't replace)
- Worktree-per-branch note: git hooks live in the shared
.git dir config; a hook installed in the main checkout fires for worktree commits too — install once per repo, not per worktree
Verify:
cat <repo>/.git/hooks/post-commit | grep -n graphify
Option B — live watch mode
graphify <target> --watch
- Code file saves → instant AST-only rebuild (no LLM, no tokens)
- Doc/image changes → notifies; run
graphify <target> --update for the LLM re-pass
- Runs in the foreground — launch it via a background Bash task or a dedicated terminal pane, NOT inline in a Claude session
- Useful for multi-agent workflows where parallel agents write code and the graph must stay current between waves
Choosing
| Situation | Pick |
|---|
| Normal repo, commit-granularity freshness is fine | Hook (A) |
| Active multi-agent session mutating files continuously | Watch (B) |
| Both installed | Fine — watch covers between-commit drift; hook covers sessions where watch isn't running |
Teardown
sed -i '' '/graphify/d' <repo>/.git/hooks/post-commit
Post-Execution Reflection
After this skill completes, reflect before closing the task:
- Locate yourself. — Find this SKILL.md's canonical path before editing.
- What failed? — Fix the instruction that caused it.
- What worked better than expected? — Promote to recommended practice.
- What drifted? — Fix any script, reference, or dependency (esp.
references/backends.md) that no longer matches reality.
- Log it. — Add a dated note to the plugin CLAUDE.md provenance with trigger, fix, and evidence.
Do NOT defer. The next invocation inherits whatever you leave behind.
