verify-fix
Pass
Audited by Gen Agent Trust Hub on Sep 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- Indirect Prompt Injection Surface: The skill is designed to process external security findings and issue reports, which represents a potential surface where instructions embedded in the input data could attempt to influence the agent's behavior.
- Ingestion points: The skill in SKILL.md processes a 'supplied security finding' and its 'supplied contents'.
- Boundary markers: The instructions do not specify explicit delimiters or markers to separate ingested content, though the strict JSON output contract acts as a structural boundary.
- Capability inventory: The skill workflow in SKILL.md involves tracing code paths and 'running reproducers,' which involves file system access and potentially shell execution.
- Sanitization: The workflow does not explicitly mention validation or filtering of the input finding data.
- Command Execution for Verification: The instructions include a step to run reproducers or regression checks to verify remediation. This is an intended functional requirement of the skill, and the risk is addressed by an explicit instruction to only perform these checks if they do not modify the repository environment.
Audit Metadata