defensive-patterns
Pass
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill contains no executable code or scripts and serves purely as a collection of architectural best practices. It advocates for security-enhancing measures such as least privilege for file access and environment scrubbing to prevent credential leakage.
- [INDIRECT_PROMPT_INJECTION]: The skill is intended to guide the agent while it processes untrusted data, such as during code reviews or when handling external I/O.
- Ingestion points: External source code and project documentation reviewed by the agent (as defined in SKILL.md).
- Boundary markers: The guidelines recommend async settlement and callback containment to establish clear execution boundaries.
- Capability inventory: The skill discusses the use of subprocess spawning and filesystem interactions (referenced in SKILL.md).
- Sanitization: The instructions explicitly mandate scrubbing sensitive credentials (keys, secrets, tokens) from the environment and using exclusive file permissions.
Audit Metadata