design-consultation
Pass
Audited by Gen Agent Trust Hub on Jun 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches the Bun runtime installation script from bun.sh, which is a well-known technology service. The skill also performs web research and captures snapshots of external sites to inform the design process.\n- [REMOTE_CODE_EXECUTION]: Executes the downloaded Bun installation script via the shell. This execution is gated by a hardcoded SHA-256 checksum verification to ensure the integrity of the downloaded file before it is run.\n- [COMMAND_EXECUTION]: Performs numerous local shell operations including managing project files, executing git commands, and running specialized design and browsing binaries located in the user's home directory. It also starts a local HTTP server to serve design comparison boards.\n- [PROMPT_INJECTION]: The skill processes untrusted data which presents an indirect prompt injection surface:\n
- Ingestion points: Reads project files such as README.md and package.json, and ingests external website data during the research phase.\n
- Boundary markers: Does not utilize explicit delimiters to isolate ingested content from its internal instructions.\n
- Capability inventory: Significant capabilities including execution of subprocesses, file modification, and network access.\n
- Sanitization: Lacks explicit validation or filtering of content retrieved from external research sources.
Audit Metadata