interrogate
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's architecture relies on interpolating untrusted data into prompts, which creates a surface for indirect prompt injection.
- Ingestion points:
SKILL.mdidentifies and packages code diffs and file contents (Step 1) and user intent strings (Step 2) to be used as inputs for the sub-agent prompts. - Boundary markers: The template in
references/reviewer-prompt.mdlacks clear delimiters or protective instructions to distinguish the code being reviewed from the reviewer's instructions, potentially allowing malicious content within the analyzed code to influence the reviewer's output. - Capability inventory: The skill can spawn multiple sub-agents, read and write to the local filesystem (output and receipt paths), and send data to external LLM providers via a dispatch launcher.
- Sanitization: No sanitization, escaping, or validation logic is defined to process the code or intent strings before they are embedded into the prompt templates.
Audit Metadata