skills-search
Pass
Audited by Gen Agent Trust Hub on Mar 1, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using
npxand the vendor-specificthinkfleet-hubCLI tool to search for, install, and verify skills. - [EXTERNAL_DOWNLOADS]: Uses
npxto dynamically download and execute the@thesethrose/skills-searchpackage and theskillsCLI from the npm registry. It also connects to the officialskills.shAPI to retrieve registry data. - [DATA_EXPOSURE]: Accesses the local file system at
~/.thinkfleetbot/workspace/.agents/skills/. This access is confined to the application's own installation and workspace directory for verification purposes. - [PROMPT_INJECTION]: The skill processes external content from the
skills.shAPI. This creates an indirect prompt injection surface where malicious metadata in a third-party skill registry entry could potentially influence agent behavior during the search or installation workflow. - Ingestion points: External API results from
https://skills.sh/api/skillsare displayed in the terminal. - Boundary markers: None identified for the API data output.
- Capability inventory: Subprocess execution via
npxand file system read vials. - Sanitization: No explicit sanitization or validation of the registry data is shown before display.
Audit Metadata