write-prd
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection by processing external data to build PRDs.
- Ingestion points: Data enters the agent context via the product argument and memory search results. The skill also permits web fetching and searching via its allowed tools.
- Boundary markers: The templates and instructions lack specific markers or system-level directives to differentiate between retrieved content and instructions, which could lead the agent to follow commands embedded in external documents.
- Capability inventory: The skill is configured with broad tool access, including Bash shell execution, file writing, and task management, which increases the potential impact of a successful injection.
- Sanitization: No sanitization or validation logic is defined for data fetched from external sources or project memory before it is interpolated into prompts or used to drive subtasks.
Audit Metadata