mdq
Fail
Audited by Gen Agent Trust Hub on Mar 13, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides instructions to download pre-compiled executable binaries from a GitHub repository ('yshavit/mdq') that is not recognized as a trusted organization or well-known technology provider.
- [COMMAND_EXECUTION]: The installation process involves executing shell commands with elevated privileges using 'sudo' to move a binary into a system directory ('/usr/local/bin/').
- [PROMPT_INJECTION]: The skill is designed to parse and filter external Markdown files, which introduces an indirect prompt injection surface where instructions hidden within the Markdown could potentially influence the agent's behavior. * Ingestion points: External Markdown files read by the 'mdq' tool. * Boundary markers: None identified. * Capability inventory: Command execution of the downloaded 'mdq' binary. * Sanitization: No sanitization of input Markdown content is mentioned.
Recommendations
- AI detected serious security threats
Audit Metadata