vs-search-tuning

Pass

Audited by Gen Agent Trust Hub on Jul 27, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill contains an indirect prompt injection surface because it processes untrusted external data (query sets) that are used as inputs for search similarity evaluation and LLM-based relevance judging.
  • Ingestion points: The search tune validate, plan, run, and compare commands ingest data from external files via the --queries flag.
  • Boundary markers: The skill instructions do not specify the use of delimiters or 'ignore' instructions to prevent the LLM judge from executing commands embedded within the query text.
  • Capability inventory: The search tune run command performs network operations (search requests and LLM API calls) and file writes (tuning artifacts).
  • Sanitization: There is no mention of sanitizing or filtering the content of the ingested query files before they are processed by the search system or LLM judge.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 27, 2026, 06:27 AM
Security Audit — agent-trust-hub — vs-search-tuning