deep-research
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze data from untrusted external sources (web pages, GitHub, etc.) to synthesize research. This creates an attack surface where instructions embedded in external content could influence the agent's behavior.\n
- Ingestion points: The
WebFetchtool is used to retrieve content from external domains (e.g., arXiv, IETF, W3C, engineering blogs) and GitHub repositories (READMEs, design docs).\n - Boundary markers: The instructions do not define specific delimiters or "ignore" commands to isolate external data from the agent's core instructions, though they emphasize cross-validation and direct quoting.\n
- Capability inventory: The skill uses data retrieval tools (
WebFetch,Context7) to gather information but does not include capabilities for file system modification, arbitrary command execution, or network exfiltration of sensitive data.\n - Sanitization: There is no explicit instruction to sanitize, filter, or validate the structure of retrieved content before it is processed by the model.
Audit Metadata