Loading ToolBench
Design-intelligent MCP server for Figma. 201 design-intelligent tools for Figma. Every tool knows typography, spacing, color, accessibility. Not a shape proxy — a design engine.
conductor-figma has 6 tools with basic schemas and descriptions, but significant gaps prevent production use. Tool names lack clear action verbs (e.g., 'get_design_craft_guide' is vague about what 'guide' means). Descriptions are present but generic (avg ~80 chars, below 194-char baseline). Input schemas exist but lack depth: most parameters have minimal type info and no validation constraints (enums, ranges, patterns). Output schemas are completely undocumented, LLMs cannot predict what these tools return. Error handling is absent; no recovery guidance. The 'design_from_prompt' tool is destructive (WRITE risk) but has no confirmation pattern or dry-run option. Parameter descriptions are sparse (e.g., 'ratio' in suggest_type_scale lists values but no explanation of what each ratio does). No tool annotations (readOnlyHint/destructiveHint) despite clear read/write distinction.
Checks WCAG contrast ratio between foreground and background colors
Generates complete Figma design from natural language prompt
Returns comprehensive design craft guide with typography, spacing, color, and accessibility best practices
Interprets design prompt and returns structured interpretation summary
Generates semantic color palette from brand color
Generates typographic scale from base size and ratio
Output schemas completely undocumented. LLMs cannot predict return types, field names, or structure. Breaks tool chaining and forces agents to guess what data is available.
Destructive tool 'design_from_prompt' (WRITE risk) lacks confirmation pattern, dry-run option, or undo capability. Agents can irreversibly modify Figma designs without safeguards.
Parameter descriptions are minimal or missing validation constraints. 'ratio' in suggest_type_scale lists enum values but no explanation of what each ratio produces. 'mode' in suggest_color_palette lacks description entirely.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 33 | <=2025-11-25 | v2 |
Tool names lack clear action verbs. 'get_design_craft_guide' is ambiguous (guide to what?). 'interpret_prompt' and 'design_from_prompt' do not distinguish intent clearly. Naming should follow verb_noun pattern (e.g., 'fetch_design_guide', 'generate_design_from_prompt').
No tool annotations (readOnlyHint/destructiveHint/idempotentHint). Server declares risk levels in metadata but does not expose them via MCP tool annotations, preventing clients from applying safety policies.