peer-review
Pass
Audited by Gen Agent Trust Hub on May 11, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill interpolates user-provided material and criteria into a prompt for a secondary analysis tool. This surface for indirect prompt injection is mitigated by architectural constraints.\n
- Ingestion points: Artifacts and review criteria are identified from user prompts in SKILL.md Step 1.\n
- Boundary markers: User-controlled content is wrapped in XML tags such as
<task>and<structured_output_contract>.\n - Capability inventory: The skill invokes the
/codex-exectool (Step 3).\n - Sanitization: The skill explicitly mandates that the
/codex-exectool be invoked in read-only mode, which prevents injected instructions from performing unauthorized file modifications or command execution.
Audit Metadata