quack
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied intent strings to route requests to specialized subagents, creating a surface for indirect prompt injection.
- Ingestion points: Untrusted user input is ingested via the '' parameter following the 'quack' command (SKILL.md).
- Boundary markers: The skill requires an explicit activation keyword ('quack') and performs normalization on the intent string to strip separators and punctuation (SKILL.md).
- Capability inventory: The skill possesses the capability to delegate tasks to subagents like 'duck-patch', 'duck-debug', and 'duck-refactor', which have the authority to suggest or perform code modifications (SKILL.md, assets/subagent-runbook.md).
- Sanitization: Normalization is limited to structural cleanup of the intent string; no deep content sanitization or safety filtering of the intent itself is performed before routing (SKILL.md).
Audit Metadata