create-verification-skill
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to 'interview' a codebase and follow instructions found in READMEs, Makefiles, or package scripts to build and run an application. A malicious repository could include deceptive instructions intended to influence the agent's behavior during the generation or verification phase.
- Ingestion points: The codebase of the repository being analyzed (SKILL.md, Section 1: 'Interview the repo, not the user').
- Boundary markers: None specified for the processing of repository documentation.
- Capability inventory: The skill allows for the execution of build commands, application launching, and UI/CLI driving (SKILL.md, Section 2: 'Launch' and 'Drive').
- Sanitization: No explicit sanitization or validation of the commands extracted from the repository is mentioned; the skill relies on the agent's assessment of the codebase.
- [COMMAND_EXECUTION]: The skill instructs the agent to identify and execute local commands for building, launching, and testing applications. While these are standard developer tasks, they involve running arbitrary code defined within the target repository.
- Evidence: 'Prefer the repo's own documented dev command (package scripts, Makefile, README quickstart)' and 'Run its own instructions end to end once: launch, doctor, drive ONE mapped feature'.
Audit Metadata