skillguard-audit
par jonathanliu811026
CommunicationRésumé
Audit agent skills for security threats before installing them.
Description
Ce que fait ce skill
Audits agent skill code for security threats before installation. Reads the full skill source, sends it to the SkillGuard API, and returns a verdict from SAFE to DANGEROUS with a numeric risk score. Supports paid audits ($0.05 USDC via x402) and a free credit option using a signed EIP-712 IOU.
Automated pre-install auditing catches threats like credential stealers that manual code review or community trust alone would miss.
Quand l'utiliser
- Checking a clawhub skill for credential stealers before installing
- Blocking installation of a skill flagged as DANGEROUS
- Scanning skill code for malware as part of an automated install workflow
- Verifying a skill's certification badge by its hash
- Assessing an agent's reputation by wallet address
Exemple de workflow
User asks: audit this skill before installing it
- 1
Reads the target skill's SKILL.md and all files under scripts/
- 2
Concatenates all code into a single string
- 3
Calls the SkillGuard API with the skill name and concatenated code
- 4
Receives a verdict and numeric risk score (0-100)
- 5
Blocks installation and reports threat details if verdict is CAUTION or DANGEROUS
Verdict SAFE (score 0) — installation proceeds, or DANGEROUS (score 73) — installation blocked with listed threat details
Prérequis
- Ethereum wallet address (needed to sign the free IOU for credit-based audits)
- USDC balance for paid audits ($0.05 per audit via x402)
Installation
clawhub install jonathanliu811026/skillguard-auditExécute la commande CLI, ou colle l'instruction dans Claude Code, Cursor, Codex ou l'agent de code que tu utilises.