create-verification-skill
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill's core workflow (Step 1: 'Interview the project') directs the agent to identify and execute build and launch commands found in the local repository (e.g.,
npm start,go run,docker-compose up). Running arbitrary commands from an untrusted checkout can lead to local code execution. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted project data, such as routes, documentation, and menu structures, to generate a new verification skill. This creates a surface where adversarial instructions embedded in the repository could manipulate the behavior of the generated skill or the agent.
- [EXTERNAL_DOWNLOADS]: The skill explicitly instructs the agent to install project dependencies ('install deps') as part of the verification setup. This can lead to the download and execution of arbitrary code from external package registries (npm, PyPI, etc.) based on the project's configuration files.
Audit Metadata