asu-discover
Pass
Audited by Gen Agent Trust Hub on Mar 13, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches project dependencies via pnpm and downloads the jinaai/jina-embeddings-v2-base-code embedding model from Hugging Face's official repository.
- [COMMAND_EXECUTION]: Executes shell commands via the Bash tool to run the CLI utility, manage package installations, and perform environment setup scripts.
- [DATA_EXFILTRATION]: Transmits user search queries and generated vector embeddings to a remote AWS API Gateway endpoint (https://x6qxzhvbd9.execute-api.us-west-2.amazonaws.com) as part of the primary RAG search functionality.
- [PROMPT_INJECTION]: Ingests untrusted content from external ASU GitHub repositories via search results returned by the API in src/client/api.ts. The skill does not implement specific boundary markers or sanitization to isolate this data from the instruction stream. The agent's capability inventory includes command execution (Bash) and file system access (Read, Glob, Grep).
Audit Metadata