Loading ToolBench
A collection of Figma design automation skills for accessibility auditing, component analysis, token management, and design system workflows
This server has 8 tools with significant definition gaps. Most tools lack visible input schemas in the source code provided, only 3 tools show explicit schema definitions (figma-import-tokens, figma-analyze-component-set, figma-annotations, figma-annotations-set, figma-arrange-component-set, figma-audit-accessibility, blame-node). Descriptions are present but generic (50-150 chars), lacking LLM-optimized guidance on WHEN to use each tool or what prerequisites exist. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) are visible. Error handling is absent from descriptions. Tool names are descriptive but some are overly long (figma-scan-code-accessibility, figma-arrange-component-set). Output schemas are not documented. The server appears to be a Figma plugin skill collection, not a traditional MCP server with explicit tool registration visible in the source.
Binary-search which Figma version introduced a target (property or child node). Finds the oldest version that still has the target using log2(N) node fetches.
Variant state machine + CSS pseudo-class mapping + per-variant diffs. Analyzes a component set to extract visual signatures, state mappings, and differences from default variant.
Read annotations from a Figma node (and optionally its children). Returns the node's annotations + available categories.
Write annotations to a Figma node. Undoable in Figma (Cmd+Z). Supports replace mode (overwrites all annotations) or append mode (keeps existing ones).
Organize variants into a labeled grid with a titled container. Positions variants in a grid layout with row/column labels and headers.
Deep a11y scorecard for one component / component set. Analyzes coverage of interactive states, focus indicators, color contrast, color-blind simulation, and WCAG compliance.
figma-scan-code-accessibility has no visible input schema in source code. Tool definition is inferred from package.json only, not from explicit MCP tool registration.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) visible in any tool definition. Agents cannot determine which tools are safe to retry or have side effects.
Descriptions lack WHEN-to-use guidance and prerequisites. E.g., figma-scan-code-accessibility says 'Standalone axe-core + JSDOM accessibility scanner' but does not explain when to call it vs figma-audit-accessibility.
Output schemas are not documented. Agents cannot plan downstream tool calls or extract required fields (e.g., what does figma-analyze-component-set return?).
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 50 | 2026-07-28+ | v2 |
Push design tokens INTO Figma as variables (create + update). Two-pass: pass 1 creates/updates literal values; pass 2 wires aliases (so targets exist first). Matches existing variables by saved Figma id → key → exact name, so re-imports don't duplicate.
Standalone axe-core + JSDOM accessibility scanner for HTML (code side of the Figma a11y workflow)
No error handling guidance in tool descriptions. If a node is not found or a variable alias target does not exist, agents have no recovery path.