Curator
Audit and curate the user's Claude Code skills and MCP servers. Detects overlapping clusters, proposes specialization/deletion, and generates router skills. Use when the user asks to clean up their setup, reduce token bloat, or has too many skills covering similar ground.
Install
$ agentstack add skill-karimjebara-curator-curator ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Links to your public security report.[](https://agentstack.voostack.com/security/report/skill-karimjebara-curator-curator) Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
curator
When invoked, run the curator audit on the user's ~/.claude/ setup and walk them through the findings.
How to use
- Run the scan via Bash:
npx -y @curator/cli scan. - Run the interactive review:
npx -y @curator/cli review. The CLI walks the user cluster-by-cluster with PASS actions pre-selected and WARN/FAIL pre-skipped — let the user drive that flow themselves. - If the user prefers to discuss before running the interactive prompt, read
~/.claude/curator/reports/latest.md, summarize cluster-by-cluster, and surface verifier WARN/FAIL notes prominently. Then either edit the markdown file to check boxes (and runcurator review --report --apply) or hand off to the interactive prompt.
What to flag
- Clusters with mean similarity above 70% — those are usually true duplicates.
- Skills with version-drift suffixes (
-v2,-enhanced, etc.) — almost always safe to delete. - MCP servers consuming more than 1500 tokens individually — high-value targets for removal or scoping.
What NOT to do
- Never apply actions without explicit user confirmation per action.
- Never delete a skill that has unique vocabulary or examples not covered elsewhere — even if curator suggested it.
- Never modify the report's structure when checking boxes; only flip
[ ]to[x].
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: KarimJebara
- Source: KarimJebara/Curator
- License: MIT
- Homepage: https://github.com/KarimJebara/Curator
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.