create-verification-skill
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to discover and execute repository-specific commands for building, launching, driving, and cleaning up the project. It also uses git commands like
git check-ignoreto verify file tracking status. - Evidence: The 'Launch', 'Doctor', 'Drive', and 'Cleanup' sections in the generated skill are populated with real commands found in the repo and executed during the 'Prove the Generated Skill' step.
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest information from the local repository (code, documentation, and configuration files) to generate a new verification skill. This creates a surface where malicious content within the analyzed repository could influence the agent's output or behavior.
- Ingestion points: The entire target repository (codebase, config files, documentation) is scanned in step 1.
- Boundary markers: Not explicitly defined for the repository discovery process, although the 'Guard' section sets scope boundaries for the agent's actions.
- Capability inventory: File system writes to
.claude/skills/,AGENTS.md, andCLAUDE.md, along with the execution of project-specific shell commands. - Sanitization: The instructions do not specify sanitization or validation routines for the data discovered within the repository.
Audit Metadata