maintain-verification-skill
Pass
Audited by Gen Agent Trust Hub on Sep 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted project source code and feature maps to generate verification recipes, creating an attack surface for indirect prompt injection.\n
- Ingestion points: Processes feature files and application source code (Steps 2 and 3 in SKILL.md).\n
- Boundary markers: No boundary markers or instructions to ignore embedded commands are specified.\n
- Capability inventory: Includes file system write capabilities within the edit scope and the execution of local scripts (Steps 4 and 5).\n
- Sanitization: No sanitization of ingested content is mentioned.\n- [COMMAND_EXECUTION]: The skill executes 'live-verification recipes' and 'harness scripts' found in the local project environment.\n
- Evidence: Steps 4 and 5 explicitly instruct the agent to exercise features live and run executable harness scripts defined in the target skill directory.\n- [DYNAMIC_EXECUTION]: The skill's execution behavior and launch model are determined at runtime by the contents of the 'Launch' section in the target skill being audited.\n
- Evidence: Step 4 describes dynamically following the verification skill's own launch model.
Audit Metadata