literature-search
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources including PubMed, arXiv, bioRxiv, and medRxiv via the Valyu semantic search API. Maliciously crafted scientific abstracts or full-text content could theoretically attempt to influence the agent's behavior.
- Ingestion points: External literature content is retrieved in
scripts/search.mjsusing thefetchAPI call tohttps://api.valyu.ai/v1/search. - Boundary markers: There are no explicit boundary markers or instructions to ignore embedded prompts within the retrieved scientific content.
- Capability inventory: The skill's scripts (
scripts/search.mjs) have the capability to read and write to the user's home directory (~/.valyu/config.json) to manage API keys, but they do not possess high-privilege capabilities such as arbitrary command execution or system-wide configuration modification. - Sanitization: No sanitization, escaping, or filtering is applied to the content returned from the scientific databases before it is presented to the agent.
Audit Metadata