figma-generate-library
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill orchestrates complex Figma Plugin API operations by generating and executing JavaScript scripts. It provides a library of helper scripts (e.g., createVariableCollection.js, bindVariablesToComponent.js) and instructs the agent to embed these into tool calls. This is a core functional requirement of the skill for interacting with the Figma canvas.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from local codebase files during the discovery phase. 1. Ingestion points: Reads and parses CSS, JSON, JavaScript/TypeScript, Swift, and Kotlin files from the codebase (documented in discovery-phase.md). 2. Boundary markers: The skill provides regex patterns and specific structural definitions for token extraction to scope the agent's interpretation of the data. 3. Capability inventory: The skill can create/modify Figma pages, variables, and components, and it can read/write a local state file. 4. Sanitization: The skill defines naming and value transformation rules but relies on the underlying LLM's guardrails for sanitizing arbitrary text from the input files.
- [SAFE]: The skill implements robust state management and idempotency. It uses a state ledger to track node IDs and provides a cleanupOrphans.js script to roll back specific changes by ID rather than relying on name-prefix matching.
- [SAFE]: The skill follows established design system best practices (Material 3, Polaris) for variable scoping, platform-specific code syntax mapping, and atomic component architecture.
Audit Metadata