Bidirectional design-to-code parity engine for Figma and MCP-compatible AI agents
DesignDiff has 6 tools with clear action-verb naming (flag_, check_, generate_, audit_). Tool descriptions are present and substantive (100-200 chars typical). However, 5 of 6 tools lack visible input schemas in the provided source, only flag_stale_mappings shows a complete Zod schema. Descriptions for check_component_parity, generate_sync_patch, audit_state_coverage, check_responsive_parity, and check_theme_parity are entirely absent from the code sample. Error handling is present (try/catch with actionable messages in flag_stale_mappings), but inconsistently applied. Output structure is documented for flag_stale_mappings (structuredContent + text), but unclear for others. Tool annotations (readOnlyHint, destructiveHint) are correctly applied to flag_stale_mappings.
Audit all Code Connect mappings in a Figma file for drift. Returns components where design has changed after the connected code file was last updated — ranked by severity, impact, and days since divergence. Reads from .figma/code-connect.json in your repo root. Returns actionable summary with specific components to fix first, estimated blast radius, and direct commands to check each one.
5 of 6 tools have no visible input schema definitions in source code. Only flag_stale_mappings shows a complete Zod schema with type constraints and descriptions. The other 5 tools (check_component_parity, generate_sync_patch, audit_state_coverage, check_responsive_parity, check_theme_parity) lack documented parameter schemas, making it impossible for LLMs to understand what inputs they accept.
5 of 6 tools have no descriptions visible in the source code. Only flag_stale_mappings has a substantive description (200+ chars). The other 5 tools are registered with no description text, violating the requirement that every tool must have a non-empty description explaining WHAT it does, WHEN to use it, and any prerequisites.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 36 | 2025-06-18+ | v2 |
Output schemas are not documented for any tool except flag_stale_mappings. LLMs cannot plan downstream tool calls or extract the right data without knowing what fields to expect. This violates the pattern requirement that tools document their return structure.
Error handling is only visible in flag_stale_mappings (try/catch with actionable error messages). The other 5 tools show no error handling in the source, risking silent failures or unhelpful error responses that leave LLMs unable to recover or self-correct.
Tool annotations (readOnlyHint, destructiveHint) are only applied to flag_stale_mappings. The WRITE-risk tool generate_sync_patch lacks a destructiveHint annotation, and the other READ-ONLY tools lack readOnlyHint annotations. This prevents agents from understanding which tools are safe to retry and which have irreversible consequences.