grill-me-thermonuclear
Pass
Audited by Gen Agent Trust Hub on Aug 11, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze user-provided plans, designs, and codebase facts. While this creates a surface for indirect prompt injection, the skill employs robust boundary logic, requiring the agent to 'Steelman, then strike' and verify assertions against the actual repository before acceptance. The risk is considered inherent to the skill's purpose and is addressed by its adversarial architecture.
- [SAFE]: The skill uses standard agent capabilities, such as repository exploration and user questioning, to fulfill its stated purpose. No obfuscated content, hardcoded credentials, or unauthorized remote code execution patterns were detected.
Audit Metadata