zone-mode
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill employs imperative tags like
<NON-NEGOTIABLE>and specific instructions such asignore this skillfor subagents. These patterns, while technically instruction overrides, are used legitimately here as scoping and enforcement mechanisms to ensure the agent follows established workflows and does not skip process steps under pressure. - [INDIRECT_PROMPT_INJECTION]: The skill acts as a process router, directing the agent to load subsequent skills based on the content of user requests. This creates a surface where the agent's interpretation of a task determines the active skill context.
- Ingestion points: User task descriptions and codebase context are used to identify relevant skills.
- Boundary markers: None are explicitly defined within the routing logic itself.
- Capability inventory: The skill has no active capabilities (file writes, network, or command execution) and only serves to name or invoke other skills.
- Sanitization: No sanitization is applied to the user's intent before the routing decision occurs.
Audit Metadata