product-interview

Pass

Audited by Gen Agent Trust Hub on Sep 4, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using git (e.g., git add, git commit, git status) to manage specification files. These operations are scoped to the local repository and are performed only after explicit user confirmation through the AskUserQuestion tool.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external data, such as external Product Requirement Documents (PRDs) and existing codebase content, which presents a potential injection surface.
  • Ingestion points: The skill reads CLAUDE.md, external PRDs, existing UI code, and prior specifications during Step 1 and Step 2.
  • Boundary markers: Logical boundaries are established through a mandatory "Contract summary" (Step 4) and a "Pre-confirm verification gate" (Step 3) that validates claims before they are committed to the spec.
  • Capability inventory: The skill can write files to meta/specs/, execute local git commands, and invoke other project skills (jtbd, generate-mockups, find-gaps).
  • Sanitization: Security is maintained through a human-in-the-loop review process (Step 6) where the user must approve the final draft before it is staged or committed.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 4, 2026, 07:14 AM
Security Audit — agent-trust-hub — product-interview