do-your-research-deep

Pass

Audited by Gen Agent Trust Hub on Jul 30, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill provides a mechanism for deep fact-checking through a structured inventory and verification process, enhancing session integrity without introducing security risks.
  • [SAFE]: The design incorporates 'blind' subagent contexts, which serves as a protective boundary to prevent bias or prompt injection from the original session context from affecting the verification results.
  • [SAFE]: The skill ingests session data and external sources for its primary function. While this presents an attack surface for indirect prompt injection, it is considered safe due to the isolation design.
  • Ingestion points: Processes session-established claims and external primary sources referenced in SKILL.md.
  • Boundary markers: Employs 'blind' subagent contexts to isolate verification tasks from the original session context.
  • Capability inventory: Dispatches fresh-context subagents to perform targeted verification tasks.
  • Sanitization: Relies on context isolation and structured reporting rather than explicit input filtering.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 30, 2026, 07:56 PM
Security Audit — agent-trust-hub — do-your-research-deep