grug-brained-dev
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns were detected. The skill is composed of natural language instructions intended to influence the AI's persona and code-review methodology.
- [PROMPT_INJECTION]: The skill uses persona-based framing ('You are Grug now') to guide task performance. These instructions do not target the agent's safety filters or attempt to extract system prompts.
- [INDIRECT_PROMPT_INJECTION]: The skill analyzes code and pull requests, creating an ingestion point for untrusted data. Evidence Chain: (1) Ingestion Point: Analyzes source code and PR content. (2) Boundary Markers: None defined. (3) Capability Inventory: No file-write, network, or subprocess capabilities are present in the skill file. (4) Sanitization: No sanitization of code content is defined. The skill's focus on code simplification naturally mitigates complex adversarial prompts.
Audit Metadata