figma-relai 0.4.0 → 0.4.2

package/README.md

@@ -4,9 +4,9 @@ English | [日本語](README.ja.md) | [中文](README.zh.md)
 
 [figma-relai.vercel.app](https://figma-relai.vercel.app)
 
-**Your AI, on the canvas.** Relai connects Claude Code, Cursor, Codex — any MCP client — to Figma, so you can read, edit, audit, and build design systems by talking to the model you already use. It works on every Figma plan, because writes go through a Figma plugin rather than the paid REST API.
+**Your AI, on the canvas.** Relai connects Claude Code, Cursor, Codex — any MCP client — to Figma, so you can read, edit, audit, and build design systems by talking to the model you already use. Writes go through a Figma plugin rather than the paid REST API, so it works on every Figma plan — and the file itself keeps your rules and rulings as you go.
 
-Relai is built on a simple position: the AI era should put designers *more* in charge, not less — taste and judgment stay with you, the labor moves to a model you already trust, with every step visible.
+Relai's position: the AI era should put designers *more* in charge, not less. Taste and judgment stay with you; the labor moves to a model you already trust, with every step visible.
 
 <img src="assets/plugin-ui.png" alt="The Relai plugin: activity feed, connection status, and a Stop button" width="380" />
 
@@ -22,7 +22,7 @@ Relai is built on a simple position: the AI era should put designers *more* in c
 
 Every command shows up in the plugin as it runs, with timing and success or failure. Click an entry to jump to that layer on the canvas. Press **Stop** if you change your mind — the rest of the batch is cancelled.
 
-Relai is how its author maintains a production design system. From one of those sessions — July 2026, numbers unedited:
+The numbers scale. Relai is how its author maintains a production design system, and this excerpt is from one of those sessions — July 2026, unedited:
 
 ```text
 audit_colors     45,509 nodes scanned · 3.9s
@@ -51,13 +51,13 @@ For Cursor, add this to `.cursor/mcp.json`:
 { "mcpServers": { "Relai": { "command": "npx", "args": ["-y", "figma-relai"] } } }
 ```
 
-**3. Ask for something.** Pairing is automatic; there is nothing to copy between windows. The `join_room` tool exists for one rare case only: two Figma files running the plugin at the same time.
+**3. Ask for something.** Pairing is automatic; there is nothing to copy between windows.
 
 ## What it's good at
 
 Understanding a design. "How is this screen put together?" gets you structure, colors, layout, and token usage in one pass, and the AI can take a screenshot to actually look at the canvas rather than guess.
 
-Bulk edits. "Translate every button label to English" or "recolor this for dark mode" become one round-trip across dozens of nodes instead of an afternoon of clicking.
+Bulk edits. "Translate every button label to English" or "recolor this for dark mode" become one round-trip across dozens of layers instead of an afternoon of clicking.
 
 Audits. `analyze_design` checks color-token coverage, auto-layout quality, component health, and accessibility (WCAG contrast, touch targets, text sizes) — or all four at once as a weighted 0–100 health score you can put in a review. It also scores agent-readiness (how prepared the file is for AI work, with top gaps), fingerprints the file's voice (its radius/spacing/type signature), and runs a ghost census for references to soft-deleted variables.
 
@@ -65,24 +65,46 @@ Design systems. Variable collections with modes, token binding, shared styles, c
 
 Everything else. `execute_figma` runs JavaScript against the Figma Plugin API directly — the same escape-hatch approach as Figma's official MCP — with a `relai.*` helper library that makes the correct pattern the shortest one, hints attached to known errors, and a lint that flags silent mistakes. If you'd rather the AI never ran code, turn it off with the plugin's "Allow code execution" toggle.
 
+## The file carries the law
+
+Rules you'd normally paste into every prompt live inside the Figma file itself, so any future session — from any MCP client, by anyone who opens the file — starts already briefed.
+
+**Conventions** are a little CLAUDE.md stored in the file: naming rules, token routing, spacing habits. The AI reads them before working.
+
+**Memory** holds your precedents. Rule once — "this gap is on purpose" — and the ruling is recorded in the file; any later edit that touches what it references gets your words attached to its result:
+
+```text
+you        "this gap is on purpose — remember it"
+file       record_precedent · saved ✓
+
+(another session, a different AI, about to "fix" the gap)
+file       precedent attached — "…is on purpose" · the edit backs off
+```
+
+The panel's Memory row lists every entry; delete any of them any time. Nothing is recorded silently.
+
+**AI no-go zones** fence off whole pages — brand masters, legal, the one that is already pixel-perfect. Writes into a guarded page are rejected before they run, with the reason on the receipt, and only you edit the guard list.
+
+**Confirmation** is one level with four stops — OPEN · RISK · BULK · ALL. At RISK, the default, only dangerous operations ask (deleting variables or styles, detach, flatten); everything else keeps moving. Set it in the panel; agents can't move it.
+
 ## You stay in control
 
-The plugin is the designer's side of the deal: a live activity feed of everything the AI does, an "AI connected" indicator that means an agent is actually paired (not just that a server is running), and a Stop button that cancels pending work. Selection and page changes you make flow back to the AI as events, so "now do the same to this one" works without re-explaining. The relay is local: file contents move only between Figma, your machine, and the AI client you already trust.
+The plugin is the designer's side of the deal: a live activity feed of everything the AI does, an AI-connected indicator that means an agent is actually paired (not just that a server is running), and a Stop button that cancels pending work. Selection and page changes you make flow back to the AI as events, so "now do the same to this one" works without re-explaining.
 
-Three dials go further when you want them. **Approvals** ("Ask before big edits") holds bulk writes and code execution until you press Approve in the panel. **Lock to selection** rejects edits outside whatever you've selected — the AI gets a clear error, not a silent pass. And **file conventions** are a little CLAUDE.md stored inside the Figma file itself: naming rules, spacing habits, do-not-touch pages — every future session, from any AI client, reads it before working. Since 0.3 the file also keeps **memory**: rulings you make ("this deviation is intent, not drift") are recorded as precedents inside the file, and any future edit that touches what they reference gets the precedent attached to its result — the AI is corrected by your own case law, in the moment. The panel's Memory row shows every entry; delete any of them any time. **AI no-go zones** guard whole pages: writes into a guarded page are rejected at dispatch, and only you edit the guard list. And confirmation is one four-stop dial — OPEN · RISK · BULK · ALL: at RISK (the default) only ghost-making and irreversible operations ask; dial to OPEN for branch workflows. The UI speaks English, 日本語, and 中文.
+**Lock to selection** rejects edits outside whatever you've selected — the AI gets a clear error, not a silent pass. The relay is local: file contents move only between Figma, your machine, and the AI client you already trust. The UI speaks English, 日本語, and 中文.
 
 ## How it works
 
 ```
 AI (any MCP client)
   ↕ stdio
-MCP server            30 tools · analysis · verification
+MCP server            32 tools · analysis · verification
   (embedded relay)    WebSocket room hub on 127.0.0.1:9055
   ↕ WebSocket
 Figma plugin          executes Plugin API calls
 ```
 
-The relay lives inside the MCP server, so there is no extra process to keep alive. When several MCP clients run at once, the first one hosts the relay and the others connect to it; if the host exits, a survivor takes over. Both sides remember their room, rejoin after restarts or sleep, and find each other without any copy-pasting.
+The relay lives inside the MCP server, so there is no extra process to keep alive. When several MCP clients run at once, the first one hosts the relay and the others connect to it; if the host exits, a survivor takes over. Both sides remember their room, rejoin after restarts or sleep, and find each other without any copy-pasting. The `join_room` tool exists for one rare case only: two Figma files running the plugin at the same time.
 
 Ports are fixed by Figma's plugin sandbox: the manifest allowlists `ws://localhost:9055–9057`, and other ports cannot work without editing `manifest.json`. That's why there is no port setting in the UI.
 
@@ -107,7 +129,7 @@ Each tool is self-describing, so the AI sees full parameter docs. The same contr
 
 ## Relai and Figma's official MCP
 
-Figma's own AI has grown fast — the official MCP server now writes to the canvas, and the Figma Design Agent collaborates right inside the editor. Both are excellent, and both are reserved for full seats on paid plans, with usage metered in AI credits and models chosen by Figma. Relai is the open-source counterpart on the other side of that line: every plan including free, whatever model and subscription you already use, everything running on your machine, and the designer holding the controls. If you have the seats, the two coexist happily — run both.
+Figma's own AI has grown fast — the official MCP server now writes to the canvas, and the Figma Design Agent collaborates right inside the editor. Both are reserved for full seats on paid plans, with usage metered in AI credits and models chosen by Figma. Relai is the open-source counterpart on the other side of that line: every plan including free, whatever model and subscription you already use, everything running on your machine, and the designer holding the controls. If you have the seats, the two coexist happily — run both.
 
 ## Optional: comments
 

package/dist/{chunk-XFLMHHMX.js → chunk-BH6KSGCN.js}

@@ -12,7 +12,7 @@ function createServer() {
   return new McpServer(
     {
       name: "Relai",
-      version: "0.4.0"
+      version: "0.4.2"
     },
     {
       instructions: `
@@ -159,6 +159,7 @@ var FIGMA_COMMANDS = [
   "audit_voice_drift",
   "audit_readiness",
   "audit_ghosts",
+  "snapshot_variables",
   // Components
   "get_local_components",
   "get_instance_overrides",
@@ -377,6 +378,31 @@ var PITFALLS = [
     pattern: null,
     hint: "",
     doc: "**Binding writes inside instances silently revert \u2014 with two escape hatches.** `setBoundVariable` on a node nested inside an INSTANCE succeeds without error, then reverts (instance roots placed directly in a main ARE writable; anything deeper is not). What does persist: reassigning the whole `fills`/`strokes` array (variable-bound paints ride inside), and range-level `setRangeBoundVariable` on text. Verify after writing; never trust the call alone."
+  },
+  {
+    pattern: null,
+    hint: "",
+    doc: "**A bound `strokeWeight` renders nothing while `strokes` is empty, and wrappers don't hug stroke thickness.** A LINE's visible weight lives entirely in its stroke paint; the frame around it keeps its own raw height/width, so the panel shows an unbound number with empty Fill/Stroke \u2014 looking broken when it isn't. Inspect the LINE child's stroke bindings, not the wrapper; and bind the wrapper's thickness-axis dimension to the same variable so nothing stays raw."
+  },
+  {
+    pattern: null,
+    hint: "",
+    doc: '**`setBoundVariable("strokeWeight", \u2026)` is a silent no-op on FRAME-family nodes.** Frames and components bind stroke weight per side \u2014 `strokeTopWeight` / `strokeRightWeight` / `strokeBottomWeight` / `strokeLeftWeight`; the uniform `strokeWeight` field only binds on stroke-geometry nodes like LINE. No error is thrown, the binding just never appears \u2014 read back `boundVariables` after writing.'
+  },
+  {
+    pattern: null,
+    hint: "",
+    doc: "**Cloning a variant escapes its component set \u2014 onto whatever page the user is viewing.** `variant.clone()` parents the copy to `figma.currentPage` as a standalone COMPONENT (not into the set) \u2014 plain nodes inside components clone to the current page too. Every `componentPropertyReferences` pointing at set-level properties is silently severed, and native SLOT children are demoted to plain FRAMEs (slot-ness needs set context). Recover by `componentSet.appendChild(clone)`, re-assigning ALL references (`visible`, `slotContentId`, \u2026) \u2014 a SLOT can be restored by cloning a surviving SLOT node from a sibling variant (SLOT clones keep their type) and relinking `componentPropertyReferences.slotContentId`. Audit every set-scoped feature after adoption; none of it comes back on its own."
+  },
+  {
+    pattern: "Component set.*existing errors",
+    hint: "The set has mismatched or duplicate variant names \u2014 every variant must carry the same axes with unique values. Rename stragglers to unique same-axis values (e.g. a temporary legacy-* value) to heal the set, then migrate instances and delete the leftovers.",
+    doc: "**Uneven variant renames/deletes poison the whole component set.** If variants end up with mismatched axis segments (some carry `axis=value`, others don't) or duplicate names, the set enters an error state where `componentPropertyDefinitions` AND every instance's `variantProperties`/`setProperties` throw (`Component set has existing errors`). Escape: rename stragglers so all variants share the same axes with unique values (temporary `legacy-*` values work), which heals the set \u2014 then migrate and delete."
+  },
+  {
+    pattern: null,
+    hint: "",
+    doc: "**`setProperties` DOES work on instances nested inside other instances** \u2014 it records as an override on the outer instance. Only variable-binding writes (`setBoundVariable`) silently revert inside instance internals. Don't skip nested instances when migrating variant or boolean properties; skip them only for binding writes."
   }
 ];
 
@@ -1178,6 +1204,7 @@ import { z as z7 } from "zod";
 var ACTIONS = {
   list_collections: ["get_variable_collections", []],
   list: ["get_variables", ["collectionId"]],
+  snapshot: ["snapshot_variables", []],
   create_collection: ["create_variable_collection", ["name", "modes"]],
   update_collection: ["update_variable_collection", ["collectionId", "name", "hiddenFromPublishing"]],
   delete_collection: ["delete_variable_collection", ["collectionId"]],
@@ -1200,7 +1227,7 @@ var ACTIONS = {
 function register6(server, sendCommand) {
   server.tool(
     "manage_variables",
-    "Design-token variables: list_collections / list (variables in a collection) / create_collection / update_collection / delete_collection / create / update / delete / add_mode / remove_mode / rename_mode / set_scopes / set_code_syntax / remove_code_syntax / create_alias / bind (variable\u2192node property) / unbind / set_node_mode / get_node_modes / tokenize (find hardcoded colors & numbers that match existing variables and bind them \u2014 fix:false to preview, fix:true to apply; scope with nodeId, default current page). Pass only the fields the action needs.",
+    "Design-token variables: list_collections / list (variables in a collection) / snapshot (compact whole-file inventory \u2014 every collection and variable name/type in one call; take one before branch merges or bulk edits, diff after) / create_collection / update_collection / delete_collection / create / update / delete / add_mode / remove_mode / rename_mode / set_scopes / set_code_syntax / remove_code_syntax / create_alias / bind (variable\u2192node property) / unbind / set_node_mode / get_node_modes / tokenize (find hardcoded colors & numbers that match existing variables and bind them \u2014 fix:false to preview, fix:true to apply; scope with nodeId, default current page). Pass only the fields the action needs.",
     {
       action: z7.enum(Object.keys(ACTIONS)),
       collectionId: z7.string().optional(),
@@ -2553,10 +2580,11 @@ function register19(server, sendCommand) {
     {
       aspect: z20.enum(["color", "layout", "components", "accessibility", "tokens", "overall", "voice", "readiness", "ghosts"]),
       nodeId: z20.string().optional().describe("Root node to analyze (default: current selection)"),
-      pageIds: z20.array(z20.string()).optional().describe("voice/ghosts: pages to scan (default: current page)")
+      pageIds: z20.array(z20.string()).optional().describe("voice/ghosts: pages to scan (default: current page; ghosts auto-chunks any count)"),
+      platform: z20.enum(["desktop", "mobile"]).optional().describe("accessibility: target-size profile \u2014 desktop 24px (default), mobile 44px")
     },
     { readOnlyHint: true },
-    async ({ aspect, nodeId, pageIds }) => {
+    async ({ aspect, nodeId, pageIds, platform }) => {
       if (aspect === "overall") {
         return runOverallAudit(aspectHandlers, nodeId);
       }
@@ -2574,7 +2602,7 @@ function register19(server, sendCommand) {
       }
       const handler = aspectHandlers.get(ASPECT_TOOL[aspect]);
       if (!handler) throw new Error(`Unknown aspect: ${aspect}`);
-      return handler({ nodeId });
+      return handler({ nodeId, platform });
     }
   );
 }
@@ -2599,7 +2627,50 @@ async function runReadiness(sendCommand) {
   });
 }
 async function runGhosts(sendCommand, pageIds) {
-  const data = await sendCommand("audit_ghosts", { pageIds }, 3e5);
+  const chunks = [];
+  if (pageIds && pageIds.length > 10) {
+    for (let i = 0; i < pageIds.length; i += 10) chunks.push(pageIds.slice(i, i + 10));
+  } else {
+    chunks.push(pageIds);
+  }
+  let data;
+  if (chunks.length === 1) {
+    data = await sendCommand("audit_ghosts", { pageIds: chunks[0] }, 3e5);
+  } else {
+    const merged = {
+      pages: [],
+      ghostRefs: 0,
+      ghostCount: 0,
+      danglingRefs: 0,
+      refsScanned: 0,
+      nodesScanned: 0,
+      remoteRefs: 0,
+      ghosts: []
+    };
+    const byId = /* @__PURE__ */ new Map();
+    for (const chunk of chunks) {
+      const part = await sendCommand("audit_ghosts", { pageIds: chunk }, 3e5);
+      merged.pages.push(...part.pages ?? []);
+      merged.ghostRefs += part.ghostRefs ?? 0;
+      merged.danglingRefs += part.danglingRefs ?? 0;
+      merged.refsScanned += part.refsScanned ?? 0;
+      merged.nodesScanned += part.nodesScanned ?? 0;
+      merged.remoteRefs += part.remoteRefs ?? 0;
+      merged.criterion = part.criterion ?? merged.criterion;
+      for (const g of part.ghosts ?? []) {
+        const entry = byId.get(g.id);
+        if (entry) {
+          entry.refCount += g.refCount;
+          entry.sampleNodes = entry.sampleNodes.concat(g.sampleNodes).slice(0, 3);
+        } else {
+          byId.set(g.id, { ...g });
+        }
+      }
+    }
+    merged.ghosts = [...byId.values()].sort((a, b) => b.refCount - a.refCount).slice(0, 30);
+    merged.ghostCount = byId.size;
+    data = merged;
+  }
   const ghosts = data.ghostRefs ?? 0;
   return standardResult({
     summary: ghosts === 0 ? `No ghost references on ${data.pages?.join(", ")} (${data.refsScanned} refs across ${data.nodesScanned} nodes; dangling: ${data.danglingRefs ?? 0}).` : `${ghosts} ghost reference(s) to ${data.ghostCount} soft-deleted variable(s) on ${data.pages?.join(", ")} \u2014 they still render but are invisible in pickers and dead on publish.`,
@@ -2882,12 +2953,13 @@ function registerAnalysisTools(server, sendCommand) {
   );
   server.tool(
     "analyze_accessibility",
-    "Check accessibility: text contrast ratios against backgrounds, touch target sizes. Use to ensure designs meet WCAG guidelines. Follow with update_node to fix contrast or sizing issues.",
+    "Check accessibility: text contrast ratios against backgrounds, interactive target sizes (WCAG 2.5.8: 24px desktop default, 44px with platform:mobile). Use to ensure designs meet WCAG guidelines. Follow with update_node to fix contrast or sizing issues.",
     {
-      nodeId: z20.string().optional().describe("Root node to analyze (default: current selection)")
+      nodeId: z20.string().optional().describe("Root node to analyze (default: current selection)"),
+      platform: z20.enum(["desktop", "mobile"]).optional().describe("Target-size profile: desktop 24px (default), mobile 44px")
     },
     { readOnlyHint: true },
-    async ({ nodeId }) => {
+    async ({ nodeId, platform }) => {
       try {
         let targetId;
         if (nodeId) {
@@ -2902,7 +2974,7 @@ function registerAnalysisTools(server, sendCommand) {
           return errorResult2("invalid_input", `Node '${targetId}' not found`, { suggestion: "Use search_nodes", tool: "search_nodes" });
         }
         const issues = [];
-        checkAccessibility(nodeInfo, null, issues);
+        checkAccessibility(nodeInfo, null, issues, platform === "mobile" ? 44 : 24);
         const data = {
           issueCount: issues.length,
           issues
@@ -3128,15 +3200,16 @@ function analyzeLayoutNode(node, issues, _counters) {
     }
   }
 }
-function checkAccessibility(node, parentBg, issues) {
-  if (node.type === "INSTANCE" || node.name?.toLowerCase().includes("button")) {
+function checkAccessibility(node, parentBg, issues, minTarget = 24) {
+  const isContainer = node.type === "PAGE" || node.type === "SECTION" || node.type === "DOCUMENT";
+  if (!isContainer && (node.type === "INSTANCE" || node.name?.toLowerCase().includes("button"))) {
     const w = node.width ?? node.absoluteBoundingBox?.width ?? 0;
     const h = node.height ?? node.absoluteBoundingBox?.height ?? 0;
-    if (w > 0 && w < 44 || h > 0 && h < 44) {
+    if (w > 0 && w < minTarget || h > 0 && h < minTarget) {
       issues.push({
         nodeId: node.id,
         nodeName: node.name,
-        issue: `Touch target too small: ${Math.round(w)}\xD7${Math.round(h)}px (minimum 44\xD744px)`
+        issue: `Target too small: ${Math.round(w)}\xD7${Math.round(h)}px (minimum ${minTarget}\xD7${minTarget}px \u2014 WCAG 2.5.8${minTarget >= 44 ? ", mobile" : ""})`
       });
     }
   }
@@ -3158,7 +3231,7 @@ function checkAccessibility(node, parentBg, issues) {
         const isLarge = fontSize !== null && (fontSize >= 24 || fontSize >= 18.66 && fontWeight >= 700);
         const required = isLarge ? 3 : 4.5;
         const ratio = calculateContrastRatio(effective, parentBg.color);
-        if (ratio < required) {
+        if (Math.round(ratio * 10) / 10 < required) {
           issues.push({
             nodeId: node.id,
             nodeName: node.name,
@@ -3288,10 +3361,11 @@ function register20(server, sendCommand) {
     "validate_design_rules",
     "Run design quality rules on a node: token coverage, auto-layout usage, naming conventions, and accessibility basics. Use after modifications or during design audits. Returns pass/fail for each rule with fix suggestions.",
     {
-      nodeId: z21.string().optional().describe("Root node to validate (default: current selection)")
+      nodeId: z21.string().optional().describe("Root node to validate (default: current selection)"),
+      platform: z21.enum(["desktop", "mobile"]).optional().describe("Target-size profile for the touch rule: desktop 24px (default), mobile 44px")
     },
     { readOnlyHint: true },
-    async ({ nodeId }) => {
+    async ({ nodeId, platform }) => {
       try {
         let targetId;
         if (nodeId) {
@@ -3363,17 +3437,19 @@ function register20(server, sendCommand) {
           });
         } catch {
         }
-        if (nodeInfo.type === "INSTANCE" || nodeInfo.name?.toLowerCase().includes("button")) {
+        const isContainerNode = nodeInfo.type === "PAGE" || nodeInfo.type === "SECTION" || nodeInfo.type === "DOCUMENT";
+        if (!isContainerNode && (nodeInfo.type === "INSTANCE" || nodeInfo.name?.toLowerCase().includes("button"))) {
+          const minTarget = platform === "mobile" ? 44 : 24;
           const w = nodeInfo.width ?? 0;
           const h = nodeInfo.height ?? 0;
-          const adequate = w >= 44 && h >= 44;
+          const adequate = w >= minTarget && h >= minTarget;
           results.push({
             rule: "touch_target_size",
             passed: adequate,
             severity: "warning",
-            message: adequate ? `Touch target: ${Math.round(w)}\xD7${Math.round(h)}px (OK)` : `Touch target: ${Math.round(w)}\xD7${Math.round(h)}px (minimum 44\xD744px)`,
+            message: adequate ? `Target size: ${Math.round(w)}\xD7${Math.round(h)}px (OK, minimum ${minTarget}px)` : `Target size: ${Math.round(w)}\xD7${Math.round(h)}px (minimum ${minTarget}\xD7${minTarget}px \u2014 WCAG 2.5.8${minTarget >= 44 ? ", mobile" : ""})`,
             nodeId: targetId,
-            fix: !adequate ? { tool: "update_node", reason: "Increase size to meet touch target minimum", args: { nodeId: targetId } } : void 0
+            fix: !adequate ? { tool: "update_node", reason: "Increase size to meet the target-size minimum", args: { nodeId: targetId } } : void 0
           });
         }
         try {
@@ -3510,18 +3586,32 @@ var moduleCategories = [
   [batch_exports, "advanced"],
   [execute_exports, "advanced"]
 ];
-function registerRoomTool(server, joinRoom) {
+function registerRoomTool(server, joinRoom, listRooms) {
   server.tool(
     "join_room",
-    "Connect to a specific Figma plugin instance by room name. Usually unnecessary \u2014 pairing is automatic when one plugin is connected. Use only when an error reports multiple plugins/rooms; the room name is shown in each plugin's UI.",
+    "Connect to a specific Figma plugin instance. Usually unnecessary \u2014 pairing is automatic when one plugin is connected. Call with no arguments to LIST the connected plugins (room + file name) instead of typing the room by hand, then call again with the room you want.",
     {
-      room: z22.string().describe("Room name shown in the Figma plugin")
+      room: z22.string().optional().describe("Room name shown in the Figma plugin. Omit to list connected rooms.")
     },
     async ({ room }) => {
       try {
         if (!room) {
+          const rooms = (await listRooms()).filter((r) => r.hasPlugin);
+          if (rooms.length === 0) {
+            return {
+              content: [{ type: "text", text: "No Figma plugin is connected. Open the Relai plugin in Figma (it connects automatically), then try again." }]
+            };
+          }
+          const lines = rooms.map(
+            (r) => `- ${r.room}${r.fileName ? ` \u2014 "${r.fileName}"` : ""}`
+          );
           return {
-            content: [{ type: "text", text: "Please provide a room name." }]
+            content: [{
+              type: "text",
+              text: rooms.length === 1 ? `One plugin connected (auto-pairing will use it):
+${lines[0]}` : `${rooms.length} plugins connected \u2014 call join_room with the room of the file you want:
+${lines.join("\n")}`
+            }]
           };
         }
         await joinRoom(room);
@@ -3593,7 +3683,7 @@ function listToolCatalog() {
 }
 
 // ../../docs/skills/figma-plugin-api.md
-var figma_plugin_api_default = '# Figma Plugin API Cheat Sheet \u2014 for `execute_figma`\n\nRules and patterns for writing code that runs in the plugin sandbox. Work in small incremental scripts and `screenshot` between steps.\n\n## Non-negotiable rules (dynamic-page mode)\n\n- **Node lookup is async**: `await figma.getNodeByIdAsync(id)` \u2014 `getNodeById` throws in this plugin.\n- **Load pages before traversal**: `await page.loadAsync()` before `page.findAll(...)` on a non-current page; `figma.currentPage` is always loaded.\n- **Load fonts before ANY text edit**: `await figma.loadFontAsync(textNode.fontName)` \u2014 if `fontName === figma.mixed`, load every range: `for (const f of textNode.getRangeAllFontNames(0, textNode.characters.length)) await figma.loadFontAsync(f)`.\n- **`figma.mixed` is a Symbol**: properties like `fontSize`, `cornerRadius`, `fills` return it when values differ across ranges/children. Check `=== figma.mixed` before using; never JSON-serialize it.\n\n## Common patterns\n\n```js\n// Selection\nconst sel = figma.currentPage.selection;          // read\nfigma.currentPage.selection = [node];             // write\n\n// Create + place\nconst frame = figma.createFrame();\nframe.resize(320, 200);                            // width/height are read-only; use resize()\nparent.appendChild(frame);                         // default parent is currentPage\n\n// Fills/strokes are ARRAYS and must be reassigned wholesale\nnode.fills = [{ type: "SOLID", color: { r: 1, g: 0.5, b: 0 } }];  // rgb 0-1, no alpha key \u2014 use opacity\nconst fills = JSON.parse(JSON.stringify(node.fills));              // clone before mutating\nfills[0].color.r = 0; node.fills = fills;\n\n// Auto-layout: set layoutMode FIRST, then padding/spacing/align\nframe.layoutMode = "VERTICAL";\nframe.itemSpacing = 8; frame.paddingTop = 16;\nchild.layoutSizingHorizontal = "FILL";             // requires PARENT with auto-layout\n\n// Components\nconst comp = await figma.importComponentByKeyAsync(key);  // team library\nconst inst = comp.createInstance();\ninst.setProperties({ Variant: "Primary" });        // throws on unknown property names\n\n// Variables\nconst collections = await figma.variables.getLocalVariableCollectionsAsync();\nconst v = figma.variables.createVariable("name", collection, "COLOR");\nv.setValueForMode(modeId, { r: 0, g: 0, b: 0 });\nnode.setBoundVariable("fills", v);                 // binding paint needs figma.variables.setBoundVariableForPaint on the paint object for fills in older APIs \u2014 prefer setBoundVariable where available\n\n// Export (returns Uint8Array)\nconst bytes = await node.exportAsync({ format: "PNG", constraint: { type: "SCALE", value: 2 } });\n```\n\n## `relai.*` helpers \u2014 the shortest path is the correct one\n\nAlongside `figma`, scripts get a `relai` object whose helpers are immune to the pitfalls below. Prefer them:\n\n```js\nawait relai.text(parent, "Hello", { font: {family:"IBM Plex Mono", style:"Regular"}, size: 14, color: {r:0,g:0,b:0} });\n                                          // loads the font BEFORE writing characters\nconst card = relai.autoLayout("VERTICAL", { name: "Card", itemSpacing: 8 });\n                                          // auto-layout frame, both axes hugging\nrelai.set(node, { layoutMode: "HORIZONTAL", width: 320, opacity: 0.9 });\n                                          // layoutMode applied first; width/height via resize()\nrelai.hug(node);                          // HUG that sticks \u2014 call after appending children\nrelai.focusRing(button);                  // clipsContent + double spread shadows that render\nconst page = await relai.page(p => p.children.some(c => c.name === "Button"));\n                                          // find pages by CONTENT \u2014 names get renamed\nconst titles = relai.query(\'FRAME[name^=Card] > TEXT\');\n                                          // CSS-like search on the current page; relai.query(node, sel) scopes to a subtree.\n                                          // Supported: TYPE, *, [name=] [name*=] [name^=] [name$=], descendant, >, comma.\n                                          // NOT supported: pseudo-classes, dot-paths, sibling combinators \u2014 use findAll.\nrelai.placeholder(section);               // construction veil so the designer sees work-in-progress\nrelai.placeholder(section, false);        // ALWAYS remove when the section is done\n```\n\nTwo important facts: **scripts are NOT atomic** \u2014 on error, changes made before the throw persist, so keep scripts small and clean up after failures; and results may carry a `warnings` array for silent mistakes the lint catches on relai-created nodes and any node ids you return (e.g. spread shadows on a non-clipping frame). Convention: **return every created/mutated node id** (`return { createdNodeIds: [...] }`) \u2014 it powers both follow-up calls and the lint.\n\n## Pitfalls that throw (or silently do the wrong thing)\n\nWhen one of these throws inside `execute_figma`, the error already carries the remedy as a `Hint:` \u2014 this list is generated from the same registry (`packages/shared/src/pitfalls.ts`).\n\n<!-- PITFALLS:START -->\n- **`not a function` errors may name no symbol** \u2014 Figma\'s VM drops the subject. Usual cause: APIs from other agent tools that don\'t exist here (`node.query()`, `figma.notify()`). Use `findAll`/`findAllWithCriteria` for search; bisect when the message names nothing.\n- **Editing text without loading its font throws** (`unloaded font`). `await figma.loadFontAsync(textNode.fontName)` first; new TextNodes default to Inter Regular; when `fontName === figma.mixed`, load every font from `getRangeAllFontNames()`.\n- **`getNodeById` throws in dynamic-page mode** \u2014 always `await figma.getNodeByIdAsync(id)`.\n- **Traversing a non-current page throws until you `await page.loadAsync()`** \u2014 `figma.currentPage` is always loaded, other pages are not.\n- **`layoutSizingHorizontal/Vertical` throw without auto-layout** \u2014 FILL requires the parent to have a `layoutMode`, HUG requires it on the node itself.\n- **`createPage()` throws on the free plan once a file has 3 pages** \u2014 reuse an existing page instead.\n- **`width`/`height` are read-only** \u2014 use `node.resize(w, h)`.\n- **`figma.mixed` is a Symbol** returned by `fontSize`, `cornerRadius`, `fills` etc. when values differ across ranges/children \u2014 check `=== figma.mixed` before use; never JSON-serialize it.\n- **Exact-name lookups are fragile** \u2014 designers rename pages and layers freely; locate nodes by type/content instead of `name ===`.\n- **Stale node ids throw `does not exist`** \u2014 nodes get deleted while you work; re-read before editing and check `node.removed`.\n- **Nodes are non-extensible** \u2014 `node.myCustomProp = x` throws `object is not extensible`; use `setPluginData` or return the data instead.\n- **Shadow `spread` renders only on shapes or frames with `clipsContent: true`** \u2014 a focus ring built from spread shadows is invisible on a non-clipping frame/component.\n- **`resize()` on an auto-layout frame silently pins that axis to FIXED**, overriding `primaryAxisSizingMode: "AUTO"`. Append children first, then set `layoutSizingHorizontal = "HUG"`; use `resize` only for the fixed cross-axis.\n- **Per-corner radius** (`topLeftRadius` \u2026) only exists on RectangleCornerMixin nodes (rectangles, frames, components) \u2014 polygons, stars and lines throw.\n- **Instance children can\'t be added or removed** \u2014 detach first, or edit the main component.\n- **Figma\'s slot feature (Convert to slot, \u21E7\u2318S) has no Plugin API** \u2014 no createSlot/convertToSlot exists (verified against typings 1.123 and at runtime, 2026-07). Scaffold a frame named "Slot", select it for the designer (navigate select), and let them press the shortcut.\n- **A converted SLOT node stops hugging** \u2014 after the designer converts a frame to a slot it carries its own fixed size. Re-assert `layoutSizingVertical = "HUG"` (and the intended horizontal mode) on the SLOT node afterwards; the API can still edit its layout props.\n- **Variantizing can silently pin the hug axis to FIXED** \u2014 observed after clone \u2192 counterAxisSizingMode/width-variable changes \u2192 combineAsVariants: the variants\' `primaryAxisSizingMode` ended up FIXED though the source hugged. Always verify sizing modes on every variant after combineAsVariants and re-assert `"AUTO"`.\n- **A backgrounded Figma window suspends the plugin iframe** \u2014 timers freeze and the socket can\'t pong, so the relay\'s staleness sweep kicks the plugin and the frozen 2s redial never fires until the window regains attention. The panel redials on visibilitychange/focus (ui.html \u22650.2.4); when a paired plugin goes missing mid-session, ask the designer to click the Figma window before assuming anything is broken.\n- **A freshly-converted slot can still report `type: "FRAME"` to the Plugin API** \u2014 the layer panel shows the slot badge while getNodeByIdAsync returns FRAME (a reload boundary was observed to fix it). Never gate automation on `type === "SLOT"`; match the frame by name and trust the designer\'s visual confirmation.\n- **Deleted variables are soft-deleted ghosts** \u2014 `getVariableByIdAsync` still resolves them with name AND working values (`Variable` has no `removed` property), so bindings to them keep rendering while being invisible in pickers and dead on publish. To test aliveness, check membership in `getLocalVariablesAsync()` (remote variables are legitimately absent \u2014 check `variable.remote`).\n- **Variables cannot be reordered via the API** (`variableIds` is read-only; the `move\u2026After` methods exist only for styles) \u2014 new variables always append last; panel order is manual-drag only. NEVER delete-and-recreate variables to reorder: bindings reference variable IDs, so every binding in the file silently dies (observed in production; the ghosts keep rendering, hiding the damage).\n- **A remote (library) text style dominates variable bindings** \u2014 on text with a remote style applied, `setBoundVariable` writes appear to succeed but snap back to the style\'s own bindings. Clear the style first (`await setTextStyleIdAsync("")`), then bind.\n- **Binding writes inside instances silently revert \u2014 with two escape hatches.** `setBoundVariable` on a node nested inside an INSTANCE succeeds without error, then reverts (instance roots placed directly in a main ARE writable; anything deeper is not). What does persist: reassigning the whole `fills`/`strokes` array (variable-bound paints ride inside), and range-level `setRangeBoundVariable` on text. Verify after writing; never trust the call alone.\n<!-- PITFALLS:END -->\n\nAlso: `resize(w, h)` throws on `w <= 0 || h <= 0`; `layoutWrap = "WRAP"` is HORIZONTAL-only, `counterAxisAlignItems = "BASELINE"` too; writing to `locked` nodes is allowed by the API but surprises designers \u2014 check first.\n\n## Return values\n\nReturn JSON-serializable data (the bridge summarizes nodes automatically and truncates >50k chars). Prefer returning `{ id, name, type }` summaries over whole nodes. `console.log` is captured and returned in `logs`.\n\n---\n\n*Tool parameter contracts are deliberately not duplicated in this document \u2014 every tool is self-describing over MCP, and `npx figma-relai docs <tool>` (or the generated `docs/manifest.json`) is the always-current reference.*\n';
+var figma_plugin_api_default = '# Figma Plugin API Cheat Sheet \u2014 for `execute_figma`\n\nRules and patterns for writing code that runs in the plugin sandbox. Work in small incremental scripts and `screenshot` between steps.\n\n## Non-negotiable rules (dynamic-page mode)\n\n- **Node lookup is async**: `await figma.getNodeByIdAsync(id)` \u2014 `getNodeById` throws in this plugin.\n- **Load pages before traversal**: `await page.loadAsync()` before `page.findAll(...)` on a non-current page; `figma.currentPage` is always loaded.\n- **Load fonts before ANY text edit**: `await figma.loadFontAsync(textNode.fontName)` \u2014 if `fontName === figma.mixed`, load every range: `for (const f of textNode.getRangeAllFontNames(0, textNode.characters.length)) await figma.loadFontAsync(f)`.\n- **`figma.mixed` is a Symbol**: properties like `fontSize`, `cornerRadius`, `fills` return it when values differ across ranges/children. Check `=== figma.mixed` before using; never JSON-serialize it.\n\n## Common patterns\n\n```js\n// Selection\nconst sel = figma.currentPage.selection;          // read\nfigma.currentPage.selection = [node];             // write\n\n// Create + place\nconst frame = figma.createFrame();\nframe.resize(320, 200);                            // width/height are read-only; use resize()\nparent.appendChild(frame);                         // default parent is currentPage\n\n// Fills/strokes are ARRAYS and must be reassigned wholesale\nnode.fills = [{ type: "SOLID", color: { r: 1, g: 0.5, b: 0 } }];  // rgb 0-1, no alpha key \u2014 use opacity\nconst fills = JSON.parse(JSON.stringify(node.fills));              // clone before mutating\nfills[0].color.r = 0; node.fills = fills;\n\n// Auto-layout: set layoutMode FIRST, then padding/spacing/align\nframe.layoutMode = "VERTICAL";\nframe.itemSpacing = 8; frame.paddingTop = 16;\nchild.layoutSizingHorizontal = "FILL";             // requires PARENT with auto-layout\n\n// Components\nconst comp = await figma.importComponentByKeyAsync(key);  // team library\nconst inst = comp.createInstance();\ninst.setProperties({ Variant: "Primary" });        // throws on unknown property names\n\n// Variables\nconst collections = await figma.variables.getLocalVariableCollectionsAsync();\nconst v = figma.variables.createVariable("name", collection, "COLOR");\nv.setValueForMode(modeId, { r: 0, g: 0, b: 0 });\nnode.setBoundVariable("fills", v);                 // binding paint needs figma.variables.setBoundVariableForPaint on the paint object for fills in older APIs \u2014 prefer setBoundVariable where available\n\n// Export (returns Uint8Array)\nconst bytes = await node.exportAsync({ format: "PNG", constraint: { type: "SCALE", value: 2 } });\n```\n\n## `relai.*` helpers \u2014 the shortest path is the correct one\n\nAlongside `figma`, scripts get a `relai` object whose helpers are immune to the pitfalls below. Prefer them:\n\n```js\nawait relai.text(parent, "Hello", { font: {family:"IBM Plex Mono", style:"Regular"}, size: 14, color: {r:0,g:0,b:0} });\n                                          // loads the font BEFORE writing characters\nconst card = relai.autoLayout("VERTICAL", { name: "Card", itemSpacing: 8 });\n                                          // auto-layout frame, both axes hugging\nrelai.set(node, { layoutMode: "HORIZONTAL", width: 320, opacity: 0.9 });\n                                          // layoutMode applied first; width/height via resize()\nrelai.hug(node);                          // HUG that sticks \u2014 call after appending children\nrelai.focusRing(button);                  // clipsContent + double spread shadows that render\nconst page = await relai.page(p => p.children.some(c => c.name === "Button"));\n                                          // find pages by CONTENT \u2014 names get renamed\nconst titles = relai.query(\'FRAME[name^=Card] > TEXT\');\n                                          // CSS-like search on the current page; relai.query(node, sel) scopes to a subtree.\n                                          // Supported: TYPE, *, [name=] [name*=] [name^=] [name$=], descendant, >, comma.\n                                          // NOT supported: pseudo-classes, dot-paths, sibling combinators \u2014 use findAll.\nrelai.placeholder(section);               // construction veil so the designer sees work-in-progress\nrelai.placeholder(section, false);        // ALWAYS remove when the section is done\n```\n\nTwo important facts: **scripts are NOT atomic** \u2014 on error, changes made before the throw persist, so keep scripts small and clean up after failures; and results may carry a `warnings` array for silent mistakes the lint catches on relai-created nodes and any node ids you return (e.g. spread shadows on a non-clipping frame). Convention: **return every created/mutated node id** (`return { createdNodeIds: [...] }`) \u2014 it powers both follow-up calls and the lint.\n\n## Pitfalls that throw (or silently do the wrong thing)\n\nWhen one of these throws inside `execute_figma`, the error already carries the remedy as a `Hint:` \u2014 this list is generated from the same registry (`packages/shared/src/pitfalls.ts`).\n\n<!-- PITFALLS:START -->\n- **`not a function` errors may name no symbol** \u2014 Figma\'s VM drops the subject. Usual cause: APIs from other agent tools that don\'t exist here (`node.query()`, `figma.notify()`). Use `findAll`/`findAllWithCriteria` for search; bisect when the message names nothing.\n- **Editing text without loading its font throws** (`unloaded font`). `await figma.loadFontAsync(textNode.fontName)` first; new TextNodes default to Inter Regular; when `fontName === figma.mixed`, load every font from `getRangeAllFontNames()`.\n- **`getNodeById` throws in dynamic-page mode** \u2014 always `await figma.getNodeByIdAsync(id)`.\n- **Traversing a non-current page throws until you `await page.loadAsync()`** \u2014 `figma.currentPage` is always loaded, other pages are not.\n- **`layoutSizingHorizontal/Vertical` throw without auto-layout** \u2014 FILL requires the parent to have a `layoutMode`, HUG requires it on the node itself.\n- **`createPage()` throws on the free plan once a file has 3 pages** \u2014 reuse an existing page instead.\n- **`width`/`height` are read-only** \u2014 use `node.resize(w, h)`.\n- **`figma.mixed` is a Symbol** returned by `fontSize`, `cornerRadius`, `fills` etc. when values differ across ranges/children \u2014 check `=== figma.mixed` before use; never JSON-serialize it.\n- **Exact-name lookups are fragile** \u2014 designers rename pages and layers freely; locate nodes by type/content instead of `name ===`.\n- **Stale node ids throw `does not exist`** \u2014 nodes get deleted while you work; re-read before editing and check `node.removed`.\n- **Nodes are non-extensible** \u2014 `node.myCustomProp = x` throws `object is not extensible`; use `setPluginData` or return the data instead.\n- **Shadow `spread` renders only on shapes or frames with `clipsContent: true`** \u2014 a focus ring built from spread shadows is invisible on a non-clipping frame/component.\n- **`resize()` on an auto-layout frame silently pins that axis to FIXED**, overriding `primaryAxisSizingMode: "AUTO"`. Append children first, then set `layoutSizingHorizontal = "HUG"`; use `resize` only for the fixed cross-axis.\n- **Per-corner radius** (`topLeftRadius` \u2026) only exists on RectangleCornerMixin nodes (rectangles, frames, components) \u2014 polygons, stars and lines throw.\n- **Instance children can\'t be added or removed** \u2014 detach first, or edit the main component.\n- **Figma\'s slot feature (Convert to slot, \u21E7\u2318S) has no Plugin API** \u2014 no createSlot/convertToSlot exists (verified against typings 1.123 and at runtime, 2026-07). Scaffold a frame named "Slot", select it for the designer (navigate select), and let them press the shortcut.\n- **A converted SLOT node stops hugging** \u2014 after the designer converts a frame to a slot it carries its own fixed size. Re-assert `layoutSizingVertical = "HUG"` (and the intended horizontal mode) on the SLOT node afterwards; the API can still edit its layout props.\n- **Variantizing can silently pin the hug axis to FIXED** \u2014 observed after clone \u2192 counterAxisSizingMode/width-variable changes \u2192 combineAsVariants: the variants\' `primaryAxisSizingMode` ended up FIXED though the source hugged. Always verify sizing modes on every variant after combineAsVariants and re-assert `"AUTO"`.\n- **A backgrounded Figma window suspends the plugin iframe** \u2014 timers freeze and the socket can\'t pong, so the relay\'s staleness sweep kicks the plugin and the frozen 2s redial never fires until the window regains attention. The panel redials on visibilitychange/focus (ui.html \u22650.2.4); when a paired plugin goes missing mid-session, ask the designer to click the Figma window before assuming anything is broken.\n- **A freshly-converted slot can still report `type: "FRAME"` to the Plugin API** \u2014 the layer panel shows the slot badge while getNodeByIdAsync returns FRAME (a reload boundary was observed to fix it). Never gate automation on `type === "SLOT"`; match the frame by name and trust the designer\'s visual confirmation.\n- **Deleted variables are soft-deleted ghosts** \u2014 `getVariableByIdAsync` still resolves them with name AND working values (`Variable` has no `removed` property), so bindings to them keep rendering while being invisible in pickers and dead on publish. To test aliveness, check membership in `getLocalVariablesAsync()` (remote variables are legitimately absent \u2014 check `variable.remote`).\n- **Variables cannot be reordered via the API** (`variableIds` is read-only; the `move\u2026After` methods exist only for styles) \u2014 new variables always append last; panel order is manual-drag only. NEVER delete-and-recreate variables to reorder: bindings reference variable IDs, so every binding in the file silently dies (observed in production; the ghosts keep rendering, hiding the damage).\n- **A remote (library) text style dominates variable bindings** \u2014 on text with a remote style applied, `setBoundVariable` writes appear to succeed but snap back to the style\'s own bindings. Clear the style first (`await setTextStyleIdAsync("")`), then bind.\n- **Binding writes inside instances silently revert \u2014 with two escape hatches.** `setBoundVariable` on a node nested inside an INSTANCE succeeds without error, then reverts (instance roots placed directly in a main ARE writable; anything deeper is not). What does persist: reassigning the whole `fills`/`strokes` array (variable-bound paints ride inside), and range-level `setRangeBoundVariable` on text. Verify after writing; never trust the call alone.\n- **A bound `strokeWeight` renders nothing while `strokes` is empty, and wrappers don\'t hug stroke thickness.** A LINE\'s visible weight lives entirely in its stroke paint; the frame around it keeps its own raw height/width, so the panel shows an unbound number with empty Fill/Stroke \u2014 looking broken when it isn\'t. Inspect the LINE child\'s stroke bindings, not the wrapper; and bind the wrapper\'s thickness-axis dimension to the same variable so nothing stays raw.\n- **`setBoundVariable("strokeWeight", \u2026)` is a silent no-op on FRAME-family nodes.** Frames and components bind stroke weight per side \u2014 `strokeTopWeight` / `strokeRightWeight` / `strokeBottomWeight` / `strokeLeftWeight`; the uniform `strokeWeight` field only binds on stroke-geometry nodes like LINE. No error is thrown, the binding just never appears \u2014 read back `boundVariables` after writing.\n- **Cloning a variant escapes its component set \u2014 onto whatever page the user is viewing.** `variant.clone()` parents the copy to `figma.currentPage` as a standalone COMPONENT (not into the set) \u2014 plain nodes inside components clone to the current page too. Every `componentPropertyReferences` pointing at set-level properties is silently severed, and native SLOT children are demoted to plain FRAMEs (slot-ness needs set context). Recover by `componentSet.appendChild(clone)`, re-assigning ALL references (`visible`, `slotContentId`, \u2026) \u2014 a SLOT can be restored by cloning a surviving SLOT node from a sibling variant (SLOT clones keep their type) and relinking `componentPropertyReferences.slotContentId`. Audit every set-scoped feature after adoption; none of it comes back on its own.\n- **Uneven variant renames/deletes poison the whole component set.** If variants end up with mismatched axis segments (some carry `axis=value`, others don\'t) or duplicate names, the set enters an error state where `componentPropertyDefinitions` AND every instance\'s `variantProperties`/`setProperties` throw (`Component set has existing errors`). Escape: rename stragglers so all variants share the same axes with unique values (temporary `legacy-*` values work), which heals the set \u2014 then migrate and delete.\n- **`setProperties` DOES work on instances nested inside other instances** \u2014 it records as an override on the outer instance. Only variable-binding writes (`setBoundVariable`) silently revert inside instance internals. Don\'t skip nested instances when migrating variant or boolean properties; skip them only for binding writes.\n<!-- PITFALLS:END -->\n\nAlso: `resize(w, h)` throws on `w <= 0 || h <= 0`; `layoutWrap = "WRAP"` is HORIZONTAL-only, `counterAxisAlignItems = "BASELINE"` too; writing to `locked` nodes is allowed by the API but surprises designers \u2014 check first.\n\n## Return values\n\nReturn JSON-serializable data (the bridge summarizes nodes automatically and truncates >50k chars). Prefer returning `{ id, name, type }` summaries over whole nodes. `console.log` is captured and returned in `logs`.\n\n---\n\n*Tool parameter contracts are deliberately not duplicated in this document \u2014 every tool is self-describing over MCP, and `npx figma-relai docs <tool>` (or the generated `docs/manifest.json`) is the always-current reference.*\n';
 
 // ../../docs/skills/design-tokens-strategy.md
 var design_tokens_strategy_default = "# Design Tokens Strategy \u2014 4-Tier Token Architecture\n\n## Token Hierarchy\n\n### Tier 1 \u2014 Core (static)\nRaw values. Never referenced directly by components.\n- Naming: `color-static/{hue}/{shade}`, `spacing-static/{value}`\n- Examples: `color-static/slate/50`, `color-static/rose/500`\n- Single mode: Value\n- hiddenFromPublishing: true\n\n### Tier 1 \u2014 Core (dynamic)\nReferences static tokens. Handles Light/Dark inversion.\n- Naming: `color-dynamic/{hue}/{shade}`\n- Light mode: shade 50 \u2192 static/50, shade 900 \u2192 static/900\n- Dark mode: shade 50 \u2192 static/900, shade 900 \u2192 static/50 (inverted)\n- 2 modes: Light / Dark\n- hiddenFromPublishing: true\n\n### Tier 2 \u2014 Themes\nSemantic tokens. Define brand personality.\n- Naming: `color/{role}/{variant}`, `typography/{style}/{property}`, `border/{type}/{size}`\n- Color roles: background, content, border, icon\n- Color variants: surface-page, surface-default, surface-subtle, surface-knockout, disabled, utility/{status}/{emphasis|subtle}\n- Typography: headline-{lg|md|sm}, title-{lg|md}, body-{lg|md|sm}[-bold], label-{lg|md|sm}, code-{md|sm}[-bold]\n- Each typography style has: font-family, font-weight, font-size, letter-spacing, line-height\n- Modes: one per brand (e.g., Brand A, Brand B)\n- description: required \u2014 explain purpose and usage constraints\n- scopes: set appropriately (TEXT_FILL for content, STROKE_COLOR for border, FRAME_FILL for background, etc.)\n\n### Tier 3 \u2014 Components\nComponent-specific tokens. Reference Tier 2 or dynamic Tier 1.\n- Naming: `{component}/color/{role}/{state}`, `{component}/size/{variant}`\n- States: default, hover, pressed, disabled, selected, selected-hover, selected-pressed, selected-disabled\n- Sizes: sm, md, lg\n- Examples: `button/color/background/default`, `button/size/md`, `navigation/color/content/selected`\n- Modes: same as Tier 2 (one per brand)\n- Not every component needs Tier 3 \u2014 only when reuse or shared decision-making justifies it\n\n## Key Rules\n- Components MUST reference tokens only \u2014 never raw hex/px values\n- Tier 1 is internal plumbing \u2014 never expose to consumers\n- Adding a new brand = adding a mode to Tier 2 and Tier 3 collections\n- Light/Dark switching is handled entirely in Tier 1 dynamic \u2014 no brand-level light/dark logic needed\n\n## Workflow\n1. `get_design_tokens` \u2192 understand existing collections and modes\n2. `manage_variables` (list) with collectionId \u2192 inspect current tokens\n3. `manage_variables` (create_collection) \u2192 new collection with modes\n4. `manage_variables` (create) \u2192 define tokens with values per mode\n5. `manage_variables` (bind) \u2192 connect tokens to node properties\n6. `manage_variables` (set_scopes) \u2192 restrict UI picker visibility\n7. `manage_variables` (set_code_syntax) \u2192 set Dev Mode code references\n\n## Verification\n- `analyze_design` (aspect: color) \u2192 find unbound colors (token coverage)\n- `validate_design_rules` \u2192 check token_coverage rule\n- `get_node_data` (detail: variables) on key nodes \u2192 confirm bindings are correct\n\n---\n\n*Tool parameter contracts are deliberately not duplicated in this document \u2014 every tool is self-describing over MCP, and `npx figma-relai docs <tool>` (or the generated `docs/manifest.json`) is the always-current reference.*\n";
@@ -3620,10 +3710,10 @@ var design_system_first_default = '# Design-system-first building\n\nThe differe
 var janitorial_cleanup_default = '# Janitorial cleanup\n\nThe chores designers postpone for months \u2014 layer renaming, token drift, spacing normalization \u2014 are where an agent earns trust fastest. These jobs are big, mechanical, and easy to verify. Do them conservatively and loudly.\n\n## Ground rules\n\n- **Preview before bulk.** Use `dryRun:true` on `batch_execute` / `set_properties`, or report findings first (`analyze_design`), and tell the designer the scale ("about 140 renames on this page \u2014 go?"). If the plugin\'s approval gate is on, big writes will ask them anyway.\n- **Scope tightly.** Work page by page (or within the designer\'s selection). Never sweep the whole file unasked.\n- **Never rename or restyle library instances\' internals** \u2014 only top-level frames/groups the team owns.\n\n## Recipes\n\n### Rename layers by content\n`Frame 427` tells nobody anything. Walk the target scope with `execute_figma` + `relai.query`, and derive names:\n- A frame whose only text child says "Submit" \u2192 `Button/Submit` (match the file\'s existing naming pattern from `get_design_system`, don\'t invent one).\n- Text nodes \u2192 their content, truncated (`Heading: Pricing plans`).\n- Skip nodes that already have deliberate names (anything not matching `/^(Frame|Group|Rectangle|Ellipse|Vector|Line) \\d+$/`).\nReturn the rename list; apply via one `batch_execute` of `rename_node` commands.\n\n### Tokenize hardcoded values\n`analyze_design aspect:"tokens"` finds hardcoded colors/numbers that visually match existing variables; `manage_variables action:"tokenize" fix:true` binds them in one pass. Report anything close-but-not-matching (deltaE just over tolerance) instead of silently snapping it.\n\n### Normalize spacing\nIn auto-layout scopes, collect `itemSpacing`/padding values; flag off-scale values (not on the file\'s 4/8pt grid or spacing variables) and propose the nearest on-scale value. Apply only after the designer confirms the mapping.\n\n### Sweep detached instances\n`analyze_design aspect:"components"` lists likely-detached instances. Offer to re-link obvious ones (matching component still exists) and just report the rest \u2014 re-attaching wrong is worse than detached.\n\n## Report format\n\nEnd with counts and jump-points: what changed, what was skipped and why, node ids for spot-checking (the designer can click entries in the plugin feed).\n\n---\n\n*Tool parameter contracts are deliberately not duplicated in this document \u2014 every tool is self-describing over MCP, and `npx figma-relai docs <tool>` (or the generated `docs/manifest.json`) is the always-current reference.*\n';
 
 // ../../docs/skills/memory-and-precedents.md
-var memory_and_precedents_default = '# File memory & precedents\n\nEvery Figma file this plugin touches can carry its own case law: **precedents** \u2014 single adjudications the designer made, stored inside the file (shared plugin data) so they reach every future session from any AI client. Conventions are the statutes; precedents are the case law. Together they are the file\'s law, and `manage_conventions action:get` returns both.\n\n## Read before you work\n\nAt the start of any session that will modify the file, call `manage_conventions action:get`. Treat what comes back the way you treat user instructions:\n\n- `content` \u2014 the conventions doc (statutes).\n- `precedents` \u2014 recent adjudications (case law), newest first, with `precedentCount` for the total.\n\nA precedent outranks your own judgment about what "looks wrong". If a precedent says a deviation is intent, do not "fix" it \u2014 and do not re-ask about it.\n\n## When to record\n\nRecord a precedent (`action:record_precedent`) when the designer **rules** on something durable. The signal is adjudication language:\n\n- "This is intentional, not drift" \u2192 `kind: "intent"`\n- "Approved \u2014 but never restructure this table again" \u2192 `kind: "decision"`\n- "Don\'t do it that way; do X instead" \u2192 `kind: "correction"`\n\nOne precedent = one sentence (\u2264280 chars), written so a future session with zero context understands the ruling. Attach `refs` \u2014 the variable IDs, node IDs, or page IDs the ruling concerns \u2014 because refs power in-band surfacing: any future write that touches them gets the precedent attached to its result automatically.\n\n**Always say in your reply that you recorded it** ("\u5DF2\u8BB0\u5165\u6587\u4EF6\u8BB0\u5FC6: \u2026"). Recording silently breaks the designer\'s trust in what the file knows.\n\n## When NOT to record\n\n- Transient, one-off choices ("make this one red for the screenshot").\n- Anything already in the conventions doc (point to it instead; propose a conventions edit if it deserves statute status).\n- Secrets, personal information, or anything about people rather than the file.\n- Your own inferences the designer has not confirmed. Precedents are the DESIGNER\'s rulings, not your observations. When unsure, ask: "\u8BB0\u4E3A\u5224\u4F8B\u5417?"\n\n## Promote rulings into enforcement\n\nWhen conventions or repeated precedents describe something ENFORCEABLE, propose the promotion in chat \u2014 never enact it silently:\n\n- "Never touch page X" (in conventions or a repeated ruling) \u2192 offer: "\u8981\u628A X \u8BBE\u4E3A AI \u7981\u533A\u5417?(manage_pages action:guard)" \u2014 set it only after the designer agrees.\n- Repeated identical adjudications ("this is intent" three times on the same token family) \u2192 offer to consolidate them into one conventions line, then remove the superseded precedents.\n\nGuards belong to the designer: never guard/unguard on your own judgment, and mention any guard change in your reply.\n\n## Reporting style as trust matures\n\nEarly sessions: report every change. Once the designer clearly stops reading full receipts (they say "\u7B80\u62A5\u5C31\u597D" or approve digests), switch to digest mode: lead with **the two changes you are least sure about** and anything a precedent or guard touched, then one line for the rest ("\u5176\u4F59 N \u5904\u6309\u89C4\u7EA6"). Full detail must always remain one question away \u2014 digest is a summary, never a omission. Never digest away errors, guard rejections, or precedent hits.\n\n## Maintenance\n\n- The designer sees and can delete every entry in the plugin panel (Memory section). Their record, their rules.\n- Memory holds 200 entries / 64KB. When recording fails with a capacity error, offer a consolidation pass: `list_precedents`, merge superseded rulings into fewer summary entries (`record_precedent` the summaries, `remove_precedent` the merged ones), preserving every still-active ruling. Show the merge plan before applying it.\n- If a precedent contradicts current reality (the token it references is gone, the ruling was reversed), ask the designer whether to update or remove it \u2014 never silently drop case law.\n';
+var memory_and_precedents_default = '# File memory & precedents\n\nEvery Figma file this plugin touches can carry its own case law: **precedents** \u2014 single adjudications the designer made, stored inside the file (shared plugin data) so they reach every future session from any AI client. Conventions are the statutes; precedents are the case law. Together they are the file\'s law, and `manage_conventions action:get` returns both.\n\nWhen a write touches something a precedent references, **the file speaks**: the ruling arrives attached to your tool result. Treat it as the designer\'s voice from a past session calling a halt \u2014 the four-act shape is always the same: they said it once, the file kept it, you came to "fix" it knowing nothing, their words stop you. Back off before you undo something intentional.\n\n## Read before you work\n\nAt the start of any session that will modify the file, call `manage_conventions action:get`. Treat what comes back the way you treat user instructions:\n\n- `content` \u2014 the conventions doc (statutes).\n- `precedents` \u2014 recent adjudications (case law), newest first, with `precedentCount` for the total.\n\nA precedent outranks your own judgment about what "looks wrong". If a precedent says a deviation is intent, do not "fix" it \u2014 and do not re-ask about it.\n\n## When to record\n\nRecord a precedent (`action:record_precedent`) when the designer **rules** on something durable. The signal is adjudication language:\n\n- "This is intentional, not drift" \u2192 `kind: "intent"`\n- "Approved \u2014 but never restructure this table again" \u2192 `kind: "decision"`\n- "Don\'t do it that way; do X instead" \u2192 `kind: "correction"`\n\nOne precedent = one sentence (\u2264280 chars), written so a future session with zero context understands the ruling. Attach `refs` \u2014 the variable IDs, node IDs, or page IDs the ruling concerns \u2014 because refs power in-band surfacing: any future write that touches them gets the precedent attached to its result automatically.\n\n**Always say in your reply that you recorded it** ("\u5DF2\u8BB0\u5165\u6587\u4EF6\u8BB0\u5FC6: \u2026"). Recording silently breaks the designer\'s trust in what the file knows.\n\n## When NOT to record\n\n- Transient, one-off choices ("make this one red for the screenshot").\n- Anything already in the conventions doc (point to it instead; propose a conventions edit if it deserves statute status).\n- Secrets, personal information, or anything about people rather than the file.\n- Your own inferences the designer has not confirmed. Precedents are the DESIGNER\'s rulings, not your observations. When unsure, ask: "\u8BB0\u4E3A\u5224\u4F8B\u5417?"\n\n## Promote rulings into enforcement\n\nWhen conventions or repeated precedents describe something ENFORCEABLE, propose the promotion in chat \u2014 never enact it silently:\n\n- "Never touch page X" (in conventions or a repeated ruling) \u2192 offer: "\u8981\u628A X \u8BBE\u4E3A AI \u7981\u533A\u5417?(manage_pages action:guard)" \u2014 set it only after the designer agrees.\n- Repeated identical adjudications ("this is intent" three times on the same token family) \u2192 offer to consolidate them into one conventions line, then remove the superseded precedents.\n\nGuards belong to the designer: never guard/unguard on your own judgment, and mention any guard change in your reply.\n\n## Reporting style as trust matures\n\nEarly sessions: report every change. Once the designer clearly stops reading full receipts (they say "\u7B80\u62A5\u5C31\u597D" or approve digests), switch to digest mode: lead with **the two changes you are least sure about** and anything a precedent or guard touched, then one line for the rest ("\u5176\u4F59 N \u5904\u6309\u89C4\u7EA6"). Full detail must always remain one question away \u2014 digest is a summary, never a omission. Never digest away errors, guard rejections, or precedent hits.\n\n## Maintenance\n\n- The designer sees and can delete every entry in the plugin panel (Memory section). Their record, their rules.\n- Memory holds 200 entries / 64KB. When recording fails with a capacity error, offer a consolidation pass: `list_precedents`, merge superseded rulings into fewer summary entries (`record_precedent` the summaries, `remove_precedent` the merged ones), preserving every still-active ruling. Show the merge plan before applying it.\n- If a precedent contradicts current reality (the token it references is gone, the ruling was reversed), ask the designer whether to update or remove it \u2014 never silently drop case law.\n';
 
 // ../../docs/skills/qa-gate.md
-var qa_gate_default = '# QA gate \u2014 the full physical, on demand\n\nHuman-triggered, review-moment audit of a file (or a set of pages). Run it when the designer says "\u4F53\u68C0" / "is this ready for review?" / before a library merge. It composes the gauges into one report with receipts.\n\n## The pass\n\nRun in this order (each is read-only):\n\n1. `manage_conventions action:get` \u2014 the law you\'ll be judging against (conventions + precedents). If precedents mark deviations as intent, they are NOT findings.\n2. `analyze_design aspect:readiness` \u2014 agent-readiness score + top gaps.\n3. `analyze_design aspect:ghosts` (pass `pageIds` for a multi-page pass) \u2014 stale references to soft-deleted variables. Any ghost count > 0 is a red flag for publish.\n4. `analyze_design aspect:voice` \u2014 the file\'s signature (radius/spacing/type, tokenized-paint rate, instance rate).\n5. `analyze_design aspect:overall` on the key screens the designer names (health score 0-100).\n6. `analyze_design aspect:tokens` \u2014 hardcoded values that match existing variables (each is one `tokenize` away from bound).\n\n## The report\n\nProduce a compact markdown report, structured exactly like this:\n\n```\n# QA gate \u2014 <file name> \u2014 <date>\nverdict: PASS | PASS WITH NOTES | BLOCK\n\nreadiness  NN/100   (top gap: \u2026)\nghosts     N refs   (0 required for publish)\nhealth     NN/100   (screens: \u2026)\ntokens     N unbound-but-matchable\nvoice      radius \u2026, spacing \u2026, \u2026% tokenized\n\n## findings (ranked)\n1. [BLOCK|NOTE] \u2026 \u2014 evidence: <numbers/nodeIds> \u2014 fix: <one tool call>\n\u2026\n\n## precedent-protected (not findings)\n- \u2026 (per precedent "\u2026", YYYY-MM-DD)\n```\n\nVerdict rules: any ghosts or health < 60 on a named key screen \u2192 BLOCK; readiness < 50 \u2192 note it but readiness alone never blocks; everything protected by a precedent is listed separately, never as a finding.\n\n## Rules\n\n- Numbers unedited: report what the tools returned, never round a failing number into a passing one.\n- Findings must carry evidence (counts, nodeIds) and a one-call fix each.\n- Offer \u2014 do not perform \u2014 fixes. The QA gate is a read-only pass; repairs are a separate, approved session.\n- If the designer wants the report in-file: create a page named `\xB7 QA Report <date>` and write the report as text there (ask first \u2014 it adds a page to their file).\n- Big files: run ghosts page-set by page-set (`pageIds`) instead of one giant pass, and say which pages were covered. Silence about coverage reads as "covered everything" \u2014 never let it.\n';
+var qa_gate_default = '# QA gate \u2014 the full physical, on demand\n\nHuman-triggered, review-moment audit of a file (or a set of pages). Run it when the designer says "\u4F53\u68C0" / "is this ready for review?" / before a library merge. It composes the gauges into one report with receipts.\n\n## The pass\n\nRun in this order (each is read-only):\n\n1. `manage_conventions action:get` \u2014 the law you\'ll be judging against (conventions + precedents). If precedents mark deviations as intent, they are NOT findings \u2014 the file already ruled. List them as "waived by precedent", never as problems.\n2. `analyze_design aspect:readiness` \u2014 agent-readiness score + top gaps.\n3. `analyze_design aspect:ghosts` (pass `pageIds` for a multi-page pass) \u2014 stale references to soft-deleted variables. Any ghost count > 0 is a red flag for publish.\n4. `analyze_design aspect:voice` \u2014 the file\'s signature (radius/spacing/type, tokenized-paint rate, instance rate).\n5. `analyze_design aspect:overall` on the key screens the designer names (health score 0-100).\n6. `analyze_design aspect:tokens` \u2014 hardcoded values that match existing variables (each is one `tokenize` away from bound).\n\n## The report\n\nProduce a compact markdown report, structured exactly like this:\n\n```\n# QA gate \u2014 <file name> \u2014 <date>\nverdict: PASS | PASS WITH NOTES | BLOCK\n\nreadiness  NN/100   (top gap: \u2026)\nghosts     N refs   (0 required for publish)\nhealth     NN/100   (screens: \u2026)\ntokens     N unbound-but-matchable\nvoice      radius \u2026, spacing \u2026, \u2026% tokenized\n\n## findings (ranked)\n1. [BLOCK|NOTE] \u2026 \u2014 evidence: <numbers/nodeIds> \u2014 fix: <one tool call>\n\u2026\n\n## precedent-protected (not findings)\n- \u2026 (per precedent "\u2026", YYYY-MM-DD)\n```\n\nVerdict rules: any ghosts or health < 60 on a named key screen \u2192 BLOCK; readiness < 50 \u2192 note it but readiness alone never blocks; everything protected by a precedent is listed separately, never as a finding.\n\n## Rules\n\n- Numbers unedited: report what the tools returned, never round a failing number into a passing one.\n- Findings must carry evidence (counts, nodeIds) and a one-call fix each.\n- Offer \u2014 do not perform \u2014 fixes. The QA gate is a read-only pass; repairs are a separate, approved session.\n- If the designer wants the report in-file: create a page named `\xB7 QA Report <date>` and write the report as text there (ask first \u2014 it adds a page to their file).\n- Big files: run ghosts page-set by page-set (`pageIds`) instead of one giant pass, and say which pages were covered. Silence about coverage reads as "covered everything" \u2014 never let it.\n';
 
 // src/prompts.ts
 var SKILLS = [
@@ -3717,4 +3807,4 @@ export {
   listToolCatalog,
   registerPrompts
 };
-//# sourceMappingURL=chunk-XFLMHHMX.js.map
+//# sourceMappingURL=chunk-BH6KSGCN.js.map