export-handoff

Pass

Audited by Gen Agent Trust Hub on Jul 13, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes pnpm action to execute local design processing tasks, including export-html, export-zip, and duplicate-design, which manage file bundling and duplication within the project directory.
  • [EXTERNAL_DOWNLOADS]: The export-design-as-figma-svg action includes functionality to fetch remote image assets from HTTP(S) sources for embedding as data URIs. Additionally, generated HTML exports include links to well-known CDN services for Tailwind CSS and Alpine.js.
  • [DATA_EXFILTRATION]: The export-coding-handoff action is designed to generate tokenized URLs and ZIP archives that provide external coding agents with access to design source files for the purpose of developer handoff.
  • [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests user-controlled design data to construct handoff documentation. 1. Ingestion points: Untrusted design data is accessed via get-design. 2. Boundary markers: The handoff prompt template lacks explicit delimiters to distinguish design data from instructions. 3. Capability inventory: Includes the ability to execute pnpm commands and write to the local file system. 4. Sanitization: No content validation or sanitization of design tokens or Alpine.js states is specified before they are interpolated into prompts.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 13, 2026, 04:51 AM
Security Audit — agent-trust-hub — export-handoff