higgsfield-to-figma

Pass

Audited by Gen Agent Trust Hub on Aug 12, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a local Python utility, scripts/plate_tools.py, to perform batch file operations including downloading plates and uploading them to Figma. The script relies entirely on Python's standard library and does not require external dependencies.
  • [EXTERNAL_DOWNLOADS]: The skill facilitates the download of AI-generated assets from Higgsfield (hosted on Amazon CloudFront) and their subsequent upload to Figma's API. Both services are well-known and professional infrastructure providers.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes structured data from Figma design templates and metadata from Higgsfield outputs, creating an attack surface for indirect prompt injection if external content contains malicious instructions designed to influence the agent's workflow.
    • Ingestion points: Figma template metadata via get_metadata (SKILL.md Step 1) and Higgsfield plate metadata/URLs (SKILL.md Step 4).
    • Boundary markers: Absent; the instructions do not include specific delimiters or warnings to ignore embedded instructions within design files or media metadata.
    • Capability inventory: Shell execution of local scripts, write-enabled Figma API access, and Higgsfield media generation.
    • Sanitization: The skill mandates a manual QA process for visual artifacts in image plates but does not implement automated sanitization or validation for text inputs interpolated into Figma layers.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 12, 2026, 09:12 PM