arxiv-search

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses a bash wrapper (scripts/search) to execute its Node.js logic. The documentation also provides a safe find command pattern for the agent to locate the script's installation path within the local environment.
  • [EXTERNAL_DOWNLOADS]: The skill makes legitimate API requests to api.valyu.ai. This is a well-known service used for the skill's primary purpose of semantic search. The network operations are limited to these official endpoints.
  • [INDIRECT_PROMPT_INJECTION]: As a search utility, the skill ingests data from external sources (arXiv preprints).
  • Ingestion points: The scripts/search.mjs script fetches article content and abstracts from the Valyu API.
  • Boundary markers: None identified in the script output; the results are returned as raw JSON.
  • Capability inventory: The skill has file system access for configuration (~/.valyu/config.json) and network access for API calls, but does not perform dynamic code execution on the ingested data.
  • Sanitization: Results are parsed as JSON and printed to standard output; no executable commands are derived from the ingested text.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 06:48 AM