maintain-verification-skill
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external project files which could contain instructions aimed at influencing the agent's maintenance logic.
- Ingestion points: Project source code, feature map documentation, and configuration files of the target verification skill.
- Boundary markers: Absent; the instructions do not provide specific methods for the agent to treat file content as untrusted data or to ignore instructions embedded within it.
- Capability inventory: Capability to execute local harness scripts, modify local files within the skill directory, and initiate Pull Requests.
- Sanitization: Absent; data from project files is interpreted by the agent and used to drive automation without a sanitization or validation layer.
- [DYNAMIC_EXECUTION]: The skill performs a cycle of modifying and then executing local automation scripts.
- Evidence: The 'Edit scope' and 'Triage' instructions guide the agent to edit harness scripts and subsequently re-run them to verify fixes, establishing a dynamic write-then-execute flow on project-local files.
Audit Metadata