research-cli
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to fetch and process content from external sources such as web pages, PDFs, and GitHub repositories. This content is inherently untrusted and could contain instructions designed to influence the agent's behavior.
- Ingestion points: Content is ingested via the
research web fetch,research pdf, andresearch web searchcommands described inSKILL.md. - Boundary markers: The instructions do not define specific delimiters or "ignore instructions" markers for the fetched content.
- Capability inventory: The skill enables execution of the
researchCLI tool to interact with external data. - Sanitization: No explicit sanitization or validation steps for the retrieved external content are mentioned in the instructions.
- [COMMAND_EXECUTION]: The skill provides instructions for the agent to execute shell commands using a specialized
researchCLI. The instructions include safety constraints, such as prohibiting the use of pipes, backgrounding, or error suppression, which limit the complexity and potential for command misuse.
Audit Metadata