pixel
Pass
Audited by Gen Agent Trust Hub on Apr 16, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill ingests untrusted image mockups and existing code, creating a surface for indirect prompt injection. Evidence: Ingestion points include user-provided PNG/JPG files and existing HTML code processed in the SCAN and VERIFY phases (SKILL.md). Boundary markers are absent in the prompts described in the precision spec. Capability inventory includes file system writes and subprocess execution of Node.js and Playwright scripts (references/visual-verification.md). No explicit sanitization of visual input is documented.
- [COMMAND_EXECUTION]: The skill documentation describes the generation and execution of local Node.js and Playwright scripts to perform visual verification and capture screenshots (references/visual-verification.md). This involves dynamic script generation and command-line execution within the local environment.
Audit Metadata