figma-generate-library

Pass

Audited by Gen Agent Trust Hub on Apr 28, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core functionality of reading and parsing external codebase files (CSS, Tailwind config, and component files) to define design system properties. \n
  • Ingestion points: Phase 0 discovery logic which ingests codebase content for token and component extraction.\n
  • Boundary markers: Absent; the instructions do not prescribe the use of delimiters or isolation protocols when handling codebase data.\n
  • Capability inventory: The agent has the ability to mutate Figma documents via use_figma and write state files to the local filesystem.\n
  • Sanitization: Absent; no validation or escaping of extracted strings is required before they are used in Figma node creation or variable naming.\n- [COMMAND_EXECUTION]: The skill utilizes dynamic script generation to orchestrate multiple Figma API operations via the use_figma tool. It instructs the agent to assemble JavaScript strings by embedding local helper scripts from the scripts/ directory and incorporating variables derived from the codebase, which constitutes a dynamic code execution surface.\n- [DATA_EXFILTRATION]: The skill implements state persistence by writing session data to /tmp/dsb-state-{RUN_ID}.json. While necessary for maintaining state across long-running workflows, this involves writing agent-controlled metadata to the host filesystem.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 28, 2026, 03:15 PM