ask-then-build

Pass

Audited by Gen Agent Trust Hub on Sep 4, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest user answers and feature ideas, which are then used to update authoritative project documentation (such as ADRs, requirements, and READMEs) and to generate a 'concise paragraph prompt' for another AI agent. This creates a pathway where malicious instructions provided by a user could be stored in a trusted file or executed by a subsequent agent in a pipeline.
  • Ingestion points: User responses to scoping questions provided in Phase 1.
  • Boundary markers: The skill lacks explicit delimiters or instructions to the LLM to treat the user-provided content as data rather than instructions when generating the final build prompt.
  • Capability inventory: The skill performs file-write operations to update repository documentation.
  • Sanitization: There is no evidence of filtering, escaping, or validation of user input before it is interpolated into project documents or the final agent prompt.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 4, 2026, 05:48 AM
Security Audit — agent-trust-hub — ask-then-build