skill-security-auditor
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill contains various prompt injection strings such as 'Ignore previous instructions', 'Act as root', and 'Skip safety checks'. These are explicitly presented within a documentation table ('What Gets Scanned') to illustrate the patterns the auditor detects in other skills. They are data to be processed, not instructions for the agent.
- [DATA_EXFILTRATION]: The skill mentions sensitive file paths (e.g., '
/.ssh/id_rsa', '/.aws') and network communication methods. These are listed as diagnostic patterns for the auditor to find in external code and do not represent active exfiltration attempts by this skill. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data (scripts and markdown files from other skills), which creates an inherent surface for indirect prompt injection.
- Ingestion points: Local directories or remote Git repositories provided to the auditing script.
- Boundary markers: None identified in the provided documentation.
- Capability inventory: Static file analysis and potential network requests for reputation checks.
- Sanitization: None described in the documentation.
Audit Metadata