reviewer-protocol
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines an orchestration workflow where a Coordinator agent follows instructions from a Reviewer agent to reassign or escalate tasks. This creates a surface where the Reviewer could potentially inject instructions that influence the Coordinator's logic beyond simple rejection. \n
- Ingestion points: Reviewer feedback, approval, and rejection comments (SKILL.md).\n
- Boundary markers: None; the instructions do not specify delimiters to isolate reviewer input from the agent's internal logic.\n
- Capability inventory: The skill empowers the agent to reassign work to different agents, spawn new agents with specific expertise, and block specific agents from contributing to artifacts.\n
- Sanitization: No validation or sanitization of the Reviewer's recommendations is mentioned.\n- [NO_CODE]: The skill is composed entirely of instructional Markdown content and does not include any scripts, libraries, or executable binaries.
Audit Metadata