initiative
Pass
Audited by Gen Agent Trust Hub on Nov 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input (product ideas, problem statements, and feature requests) and incorporates it into persistent markdown artifacts stored within the project repository.
- Ingestion points: User input is gathered via the
ask user question toolduring the initial information gathering phase (Step 1) and the Discovery Interview (Step 6.5). - Boundary markers: The skill instructions do not define explicit delimiters, headers, or "ignore embedded instructions" warnings when interpolating user-provided text into the generated initiative artifact.
- Capability inventory: The skill possesses the capability to write files to the filesystem (
initiatives/directory) and can trigger or suggest subsequent skill executions such as$work-item --mode=refineor$task-track. - Sanitization: There is no evidence of sanitization, escaping, or validation logic applied to the user-provided content before it is written to the persistent markdown files, creating a potential vector for stored prompt injection that could affect downstream processes.
Audit Metadata