find-best-skill
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes third-party instructions from SKILL.md files (local or remote), which is a common surface for indirect prompt injection. The skill mitigates this risk by executing instructions within isolated subagents using the Agent tool, wrapping content in
<skill_instructions>boundary markers, and performing a dependency scan to warn users about potential shell commands or script executions within the analyzed skill. - [EXTERNAL_DOWNLOADS]: Fetches skill definitions from user-provided URLs via WebFetch, specifically supporting GitHub and raw.githubusercontent.com. This behavior is transparent, core to the skill's comparison logic, and targets well-known services.
- [COMMAND_EXECUTION]: Uses Bash for environment resolution, directory management, and ensuring test artifacts are ignored by Git. It employs proper sanitization of user-controlled variables (skill names) by stripping non-alphanumeric characters and lowercasing them before inclusion in shell commands, preventing command injection.
- [DYNAMIC_EXECUTION]: Orchestrates the execution of complex comparison workflows using the Agent tool. This dynamic execution is the intended primary function of the skill and is implemented with isolation in mind, preventing cross-contamination between compared skills.
Audit Metadata