ask-nodumb
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-provided product ideas, observations, and project documentation, which constitutes an ingestion surface for untrusted data.
- Ingestion points: The agent reads user inputs to populate structured analysis sections such as the 'Idea Dump' and 'Fact/Trigger' as defined in the methodology.
- Boundary markers: The methodology employs strict structural delimiters (e.g.,
[JOB],[PAIN],[CAUSE]) to separate external data from the agent's internal analysis logic. - Capability inventory: The instructions focus entirely on analysis and consultation; the skill contains no instructions for writing to the filesystem, executing shell commands, or making unauthorized network requests.
- Sanitization: The skill mandates that the agent clearly label interpretations and hypotheses as distinct from confirmed facts, providing a conceptual layer of protection against the model misinterpreting external data as definitive instructions.
Audit Metadata