maintain-verification-skill
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to read and analyze arbitrary source code files and feature definitions within a project. If these files contain malicious instructions or specially crafted comments, they could attempt to influence the agent's behavior during the 'Source wave', 'Reconcile', or 'Triage' phases.
- Ingestion points: Project source code files (Step 2), feature files (Step 2), and existing verification skill files (Step 0).
- Boundary markers: No explicit delimiters or instructions to ignore embedded commands within the analyzed source code are specified.
- Capability inventory: The skill allows the agent to execute commands ('Live pass' in Step 4), perform file writes within a scoped directory (Step 5), and create pull requests (Step 6).
- Sanitization: No sanitization of ingested source code or external feature descriptions is defined before they are processed by the subagents or the coordinator.
Audit Metadata