Audit AI agent skills and prompts for safety before deployment.
npx skills add https://github.com/Ascend/agent-skills --skill skill-auditor-ascend
Please help me install this Agent Skill. Skill: skill-auditor Source: https://github.com/Ascend/agent-skills/tree/main/skills/skill-auditor Command: npx skills add https://github.com/Ascend/agent-skills --skill skill-auditor-ascend
Vet AI agent skills and prompts before deployment to prevent unsafe capabilities, prompts leakage, and data exposure.
Run the auditor on a target Skill's SKILL.md to produce an actionable audit report.
High-intent search queries and answers about installing and using this skill.
Vetting AI agent skills for security risks involves running a structured audit protocol that checks metadata, permissions, dependencies, and prompt injection vulnerabilities. The auditor generates an AGENT AUDIT REPORT with verdicts, red flags, and deployment guidance.
Prompt injection scanning for AI skills detects malicious instructions embedded in agent prompts or configurations that could cause unsafe capabilities or data leakage. The auditor enforces a fixed six-step protocol to identify these exfiltration risks.
Checking AI skill permissions and dependencies for compliance issues is a core step in the audit protocol. The auditor analyzes permission scopes and reviews dependencies to identify typosquatting and exfiltration risk before generating deployment guidance.
Detecting typosquatting in AI agent skill metadata is performed through validation checks during the audit protocol. The auditor examines metadata configurations to flag suspicious naming similarities, preventing the deployment of malicious or unsafe skills.
Limitations of automated security auditing for AI prompts include its fixed six-step protocol scope, focusing strictly on metadata, permissions, dependencies, and prompt injection scans. It generates guidance but requires human review for final deployment decisions.
caoronglinAnalyze AI agent skills for security risks and installation verdicts.
apacheAudit GitHub Actions workflows for security vulnerabilities and remediation suggestions.
Tashima-TarshValidate DISHA production deployments against security, compliance, and CI/CD requirements.
jpoutrinSecure MCP servers and multi-agent pipelines against prompt injection and unauthorized access.
