awesome-privacy-audit
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill analyzes untrusted source code and configuration files, creating an attack surface where malicious instructions embedded in the audited data could attempt to influence the agent's logic.\n
- Ingestion points: The skill reads schemas, models, API contracts, forms, and logs (SKILL.md).\n
- Boundary markers: There are no specific instructions to treat ingested data as potentially hostile or to use delimiters to isolate it from the agent's instructions.\n
- Capability inventory: The skill is restricted to read-only operations for auditing and reporting purposes.\n
- Sanitization: No methods for sanitizing or escaping content from the audited files are defined.
Audit Metadata