foundation-build-risk-review
Pass
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted user data, specifically product ideas, feature requests, or scope changes.
- Ingestion points: The agent ingests raw situation briefs and feature requests as shown in the
SKILL.mdusage instructions andevals/output-scenarios/ai-meeting-summarizer.md. - Boundary markers: There are no explicit instructions for the agent to use delimiters or ignore embedded instructions within the user-provided text (e.g., a user could theoretically include instructions to override the build verdict).
- Capability inventory: The skill instructions in
SKILL.mdimpose a "hard gate" explicitly forbidding the agent from writing code, scaffolding projects, recommending stacks, or designing implementation. No network or file-system-writing tools are utilized. - Sanitization: The prompt does not provide mechanisms for sanitizing or escaping the user's input before interpolation into the analysis logic.
Audit Metadata