codebase-to-demo
Fail
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill instructions explicitly direct the agent to clone repositories from arbitrary URLs and subsequently launch the applications locally using identified entry point commands (such as npm start, Makefile, or pyproject.toml) to capture screenshots. This process facilitates the execution of untrusted code from remote sources on the agent's host environment.\n- [COMMAND_EXECUTION]: The skill relies on shell commands for cloning git repositories and managing the execution of the project being analyzed.\n- [EXTERNAL_DOWNLOADS]: The skill performs network requests to 'exchangerate.host' to retrieve currency data and utilizes the Google Fonts CDN for output styling.\n- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes and analyzes untrusted external codebases and documentation (Ingestion Points). It lacks explicit boundary markers or sanitization for the ingested content and possesses significant shell execution and network capabilities (Capability Inventory), allowing malicious instructions embedded in a repository to potentially hijack the agent's behavior.
Recommendations
- AI detected serious security threats
Audit Metadata