recipe-front-design
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill captures untrusted user input via the $ARGUMENTS variable and interpolates it directly into the prompts for various sub-agents. \n- Ingestion points: The $ARGUMENTS input is assigned to variables like
requirement_contextandrequirements_verbatimin SKILL.md which are then passed to multiple sub-agents.\n- Boundary markers: There are no delimiters or 'ignore embedded instructions' warnings used when passing these requirements to tools, allowing potentially malicious instructions in the input to be interpreted by sub-agents.\n- Capability inventory: The sub-agents invoked (e.g.,dev-workflows-frontend:codebase-analyzer,dev-workflows-frontend:ui-spec-designer) possess broad capabilities for reading codebase structure and generating implementation designs.\n- Sanitization: The skill performs mechanical extraction of values without any escaping or validation before invocation, creating a surface where malicious instructions in the requirements could influence the behavior of downstream agents.
Audit Metadata