observability-audit
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data which could potentially contain malicious instructions.
- Ingestion points: Accesses repository files, changed components, and REPO_PROFILE.json.
- Boundary markers: The instructions do not define specific delimiters for untrusted input.
- Capability inventory: The skill operates in mode: write, allowing it to modify files in the repository.
- Sanitization: It includes negative constraints to prevent logging secrets or sensitive payloads, but lacks explicit sanitization for instructions embedded within the data it processes.
Audit Metadata