devil-advocate
Pass
Audited by Gen Agent Trust Hub on Sep 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is a pure instruction set (system prompt) with no accompanying scripts or executable code. It defines a persona for critiquing proposals.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process user-provided ideas and project context via search tools. 1. Ingestion points: User-supplied proposals and external project files accessed via the Read tool (SKILL.md). 2. Boundary markers: None explicitly defined to separate instructions from analyzed data. 3. Capability inventory: The skill is restricted to read-only tools (Read, Glob, Grep) as specified in the YAML frontmatter. It has no capabilities for network exfiltration or file modification across its instructions. 4. Sanitization: No input sanitization logic is present, but the lack of write or network capabilities effectively mitigates the impact of potential injection from processed data.
Audit Metadata