crossframe-review

Pass

Audited by Gen Agent Trust Hub on Jul 5, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill's primary function is to audit and review content generated by other AI components. It does not perform network requests, access sensitive system directories, or execute arbitrary code.
  • [SAFE]: Interactions with other skills are managed through standard relative file path references (e.g., ../crossframe/SKILL.md), allowing the agent to maintain context across a multi-skill workflow without elevating privileges.
  • [SAFE]: The instructions contain robust quality gates designed to detect and flag common AI failure modes, such as hallucinated citations ('Test 5: Fake Citation'), personality-based judgment ('Test 4: Character Judgment'), and the use of low-quality evidence ('AI Compliance Theater').
  • [SAFE]: No obfuscation, prompt injection attempts, or persistence mechanisms were found. The use of Chinese for the canonical semantic layer is clearly documented and intended for the skill's specific reasoning framework.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 5, 2026, 08:36 AM
Security Audit — agent-trust-hub — crossframe-review