maintain-verification-skill
Pass
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and analyzes project source code and feature definition files to generate verification recipes. Maliciously crafted content or comments within the audited source code could potentially influence the sub-agents' summaries or the coordinator's reconciliation logic.
- Ingestion points: Product source code files and local feature files (e.g., in
features/). - Boundary markers: Absent. The instructions do not specify the use of delimiters or 'ignore' instructions for external source content.
- Capability inventory: Spawning sub-agents, writing to the skill directory (
SKILL.md,features/), and executing harness scripts. - Sanitization: Absent. The skill does not explicitly describe sanitizing or validating the content read from the source code.
- [DYNAMIC_EXECUTION]: The skill instructions involve spawning multiple sub-agents ('parallel source readers') to handle feature analysis concurrently. It also specifies the execution of local 'harness scripts' during the live verification phase.
- [COMMAND_EXECUTION]: The skill is designed to 'drive' the application (servers, UIs, or CLIs) using a harness, which involves executing commands to exercise and verify features.
Audit Metadata