search-specialist

Pass

Audited by Gen Agent Trust Hub on Aug 8, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill instructs the agent to ingest and parse untrusted external content from the web, creating a vulnerability to indirect prompt injection. 1. Ingestion points: Instructions in SKILL.md direct the agent to 'Extract full content from promising results' and 'Parse structured data from pages'. 2. Boundary markers: The skill does not specify the use of delimiters or 'ignore instructions' markers when processing external content. 3. Capability inventory: The agent has the capability to search, fetch, and synthesize data from arbitrary web domains. 4. Sanitization: There are no requirements for sanitizing or validating external input before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 8, 2026, 11:36 AM
Security Audit — agent-trust-hub — search-specialist