psd-json-preview

Pass

Audited by Gen Agent Trust Hub on Mar 12, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill performs its stated function of generating code from design data. No evidence of credential theft, network exfiltration, or persistence mechanisms was found.
  • [PROMPT_INJECTION]: The skill ingests untrusted data from an external JSON file and interpolates it into generated code, creating a surface for indirect prompt injection. While core text content is escaped, some HTML attributes in the preview generator lack sanitization.
  • Ingestion points: Design data enters the context via the --json file path argument in scripts/generate_preview.py.
  • Boundary markers: None. The script does not utilize delimiters or specific instructions to prevent the execution of embedded commands in layer names.
  • Capability inventory: The skill has file system write access to the specified output directory to create HTML, CSS, and component files.
  • Sanitization: The React and Vue generators correctly use html.escape. However, the HTML preview generator in scripts/generators/html_generator.py interpolates layer names into aria-label attributes without escaping, which could lead to XSS if a malicious JSON file is processed.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 12, 2026, 03:56 PM