coordinator-response-mode
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a
Lightweight Spawn Templateused to generate system instructions for sub-agents by interpolating untrusted user input. - Ingestion points: External data such as
{specific task description}and{current user name}are embedded directly into the prompt inSKILL.md. - Boundary markers: The template lacks explicit delimiters (such as XML tags, triple quotes, or specific block markers) to separate the system instructions from the interpolated user-controlled task description.
- Capability inventory: The resulting agents are configured as
general-purposeand are explicitly instructed to use tools likememory.write,squad_decide, andsquad_state_writeto modify project state. - Sanitization: No sanitization, escaping, or validation logic is defined for the input variables before they are placed into the prompt context.
Audit Metadata