adversarial-review

Pass

Audited by Gen Agent Trust Hub on Sep 5, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [Indirect Prompt Injection Surface]: The skill processes user-supplied designs through $ARGUMENTS and local files. While this creates a surface for indirect prompt injection if a design document contains malicious instructions, the skill's specific role is to act as a critic and decompose designs from multiple perspectives, which serves as a natural mitigation against accidental obedience to embedded instructions.- [Local File Access]: The instructions allow the agent to read file paths provided by the user to analyze designs. This is a primary function of the skill and relies on user-provided context or explicit commands to access specific files within the project environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 5, 2026, 04:27 AM
Security Audit — agent-trust-hub — adversarial-review