matt-grill-questionnaire
Pass
Audited by Gen Agent Trust Hub on Aug 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to automatically open a generated HTML questionnaire in the user's default browser. It specifies platform-specific commands including
Start-Process,open, andxdg-open, and explicitly mandates quoting the path to prevent command injection.- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it reads repository files, git status, and source code to populate the questionnaire. Evidence chain: 1. Ingestion points: Repository instructions, source code, and specifications (SKILL.md Step 1). 2. Boundary markers: Absent. 3. Capability inventory: Local file system writes to temp directories (Step 3) and shell execution for browser opening (Step 6). 4. Sanitization: Path quoting for shell commands is required (Step 6).
Audit Metadata