azhua-skill-vetter
bởi fatfingererr
Git & GitHubTóm tắt
Security-first skill vetting for AI agents.
Mô tả
Kỹ năng này làm gì
A structured vetting protocol for AI agent skills before installation. Scans for red flags like credential harvesting, obfuscated code, and unauthorized network calls. Produces a risk classification and a formatted report with a clear install verdict.
Replaces ad-hoc manual inspection with a repeatable checklist that catches common attack patterns and forces explicit risk classification before any code runs.
Khi nào nên dùng
- Reviewing a ClawdHub skill before adding it to an agent
- Auditing a GitHub-hosted skill shared by another agent
- Checking whether a skill requests more permissions than its stated purpose requires
- Evaluating skills from unknown or low-reputation authors
- Producing a documented vetting record for team review before a high-risk install
Quy trình ví dụ
User asks: Vet this skill from GitHub before I install it
- 1
Fetch repo metadata to check stars, forks, and last updated date
- 2
Read all files in the skill directory, starting with SKILL.md
- 3
Scan code for red flags: external curl calls, credential file access, base64 decode, eval with external input
- 4
Evaluate permission scope against the skill's stated purpose
- 5
Classify risk level as LOW, MEDIUM, HIGH, or EXTREME
SKILL VETTING REPORT with red flags listed, permissions catalogued, risk level assigned, and a SAFE TO INSTALL or DO NOT INSTALL verdict
Cài đặt
clawhub install fatfingererr/azhua-skill-vetterChạy lệnh CLI, hoặc dán prompt vào Claude Code, Cursor, Codex hay coding agent bạn đang dùng.