research-deep

Pass

Audited by Gen Agent Trust Hub on Aug 17, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: No malicious patterns, obfuscation, or unauthorized commands were identified. The skill is entirely composed of prompts for structuring research tasks.- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process data from external research sources, which is a standard surface for indirect prompt injection. 1. Ingestion points: Results from focused searches and user research questions (SKILL.md). 2. Boundary markers: Absent. 3. Capability inventory: None (no code or tools are included). 4. Sanitization: Not present. Given the research context, this surface is consistent with the skill's purpose.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 17, 2026, 11:52 AM
Security Audit — agent-trust-hub — research-deep