google-labs-code · davideast · Jun 1, 2026 · May 29, 2026 · Jun 1, 2026 · Jun 1, 2026
diff --git a/.agents/skills/repo-review/SKILL.md b/.agents/skills/repo-review/SKILL.md
@@ -14,38 +14,47 @@ compatibility: Requires GitHub CLI (gh) authenticated and installed.
 This skill performs a "deep read" of a repository. It avoids superficial listings by first understanding the project's architecture and then evaluating open items against that context.
 
 ## Prerequisites
-* **Tooling:** The agent must have access to the `gh` CLI.
-* **Auth:** The user must be authenticated (`gh auth status`).
+
+- **Tooling:** The agent must have access to the `gh` CLI.
+- **Auth:** The user must be authenticated (`gh auth status`).
 
 ## Instructions
 
 ### Step 1: Initialize the Mental Model
+
 Before looking at tasks, understand the environment. Run these commands in sequence:
+
 1. **Metadata:** `gh repo view <owner/repo> --json name,description,stargazerCount,forkCount`
 2. **Structure:** `gh api repos/<owner/repo>/contents/`
 3. **Identity:** Read the project manifest (e.g., `package.json`, `go.mod`, `pyproject.toml`) and `README.md`.
 4. **Entry Point:** Read the primary entry file (e.g., `index.ts`, `main.go`) to identify the Public API surface.
 
 ### Step 2: Retrieve State & Delta
+
 Fetch current items and recent changes:
+
 1. **Issues:** `gh issue list --repo <owner/repo> --state open --json number,title,body,author,createdAt,updatedAt,labels,comments`
 2. **PRs:** `gh pr list --repo <owner/repo> --state open --json number,title,body,author,createdAt,updatedAt,labels,additions,deletions,changedFiles,headRefName`
 3. **24h Delta:** `gh issue list --state closed --since 24h` and `gh pr list --state merged --since 24h`.
 
 ### Step 3: Deep Analysis Logic
+
 For every issue and PR, do not just summarize the text. Perform a logic check:
-* **Issues:** Use `gh api repos/<owner/repo>/contents/<path>` to inspect the code mentioned in the report.
-* **PRs:** Use `gh pr diff <number>` to review actual implementation. Evaluate if the code follows the patterns found in Step 1.
-* **Flags:**
-    * 🚨 **SECURITY:** Scan for credentials, auth bypass, or injection keywords.
-    * 🤝 **EXTERNAL:** Prioritize PRs from non-maintainers.
-    * ✅ **READY:** Verify mergeability with `gh pr view <number> --json mergeable,statusCheckRollup`.
+
+- **Issues:** Use `gh api repos/<owner/repo>/contents/<path>` to inspect the code mentioned in the report.
+- **PRs:** Use `gh pr diff <number>` to review actual implementation. Evaluate if the code follows the patterns found in Step 1.
+- **Flags:**
+  - 🚨 **SECURITY:** Scan for credentials, auth bypass, or injection keywords.
+  - 🤝 **EXTERNAL:** Prioritize PRs from non-maintainers.
+  - ✅ **READY:** Verify mergeability with `gh pr view <number> --json mergeable,statusCheckRollup`.
 
 ### Step 4: Pattern Synthesis
+
 Group items into high-level insights:
-* **Bug Clusters:** Identify if 3+ issues share a root cause in a specific module.
-* **Hot Modules:** Flag files that appear in multiple open items.
-* **Client Patterns:** Group issues by consumer (e.g., "VS Code users are reporting X").
+
+- **Bug Clusters:** Identify if 3+ issues share a root cause in a specific module.
+- **Hot Modules:** Flag files that appear in multiple open items.
+- **Client Patterns:** Group issues by consumer (e.g., "VS Code users are reporting X").
 
 ## Output Template
 
@@ -54,22 +63,28 @@ Group items into high-level insights:
 **Snapshot:** {N} Issues ({+X} since yesterday) | {M} PRs ({+Y} since yesterday)
 
 ### ⚡ Suggested Actions Today
+
 1. **[Priority] #{Num}:** {One-line reason}
 2. **[Priority] #{Num}:** {One-line reason}
 
 ---
 
 ### 📂 Open Issues Deep Dive
+
 #### [Flag] #{Num} — {Title}
-* **Context:** Plain-English explanation of the "why" and "where" in the codebase.
-* **Key Evidence:** > {Blockquote of critical user text}
-* **Resolution Path:** {Quick fix/Design decision/Blocked}
+
+- **Context:** Plain-English explanation of the "why" and "where" in the codebase.
+- **Key Evidence:** > {Blockquote of critical user text}
+- **Resolution Path:** {Quick fix/Design decision/Blocked}
 
 ### 🔀 Pull Request Analysis
+
 #### [Flag] #{Num} — {Title}
-* **Impact:** Summary of files changed and architectural fit.
-* **Quality Note:** Honest assessment of implementation and CI status.
+
+- **Impact:** Summary of files changed and architectural fit.
+- **Quality Note:** Honest assessment of implementation and CI status.
 
 ### 🔍 Patterns & Health
-* **Hot Modules:** `{list/of/files}`
-* **Release Cadence:** Last release was `{days}` ago.
+
+- **Hot Modules:** `{list/of/files}`
+- **Release Cadence:** Last release was `{days}` ago.
diff --git a/.agents/skills/stitch-sdk-bug-bash/SKILL.md b/.agents/skills/stitch-sdk-bug-bash/SKILL.md
@@ -12,48 +12,53 @@ This skill provides a framework and instructions for finding bugs in the Stitch
 ## The Mindset: Adversarial Exploration
 
 When using this skill, do not just verify that the SDK works. Try to break it!
--   Pass invalid or boundary parameters.
--   Attempt operations on deleted or stale handles.
--   Simulate unexpected API responses if possible or find edge cases where projection might fail.
+
+- Pass invalid or boundary parameters.
+- Attempt operations on deleted or stale handles.
+- Simulate unexpected API responses if possible or find edge cases where projection might fail.
 
 ---
 
 ## Surface Areas to Cover
 
 ### 1. Root & Initialization (`Stitch`)
--   **Zero Config**: Verify the singleton works without explicit config if `STITCH_API_KEY` is present.
--   **Invalid Config**: Pass an empty API key or invalid base URL to `StitchToolClient` and verify that the first call fails with a clear authentication or connection error, not a generic noise error.
+
+- **Zero Config**: Verify the singleton works without explicit config if `STITCH_API_KEY` is present.
+- **Invalid Config**: Pass an empty API key or invalid base URL to `StitchToolClient` and verify that the first call fails with a clear authentication or connection error, not a generic noise error.
 
 ### 2. Project Lifecycle (`Project`)
--   **Handle Creation**: Verify that `stitch.project('invalid-id')` does not throw (lazy instantiation) but the first call on it fails safely.
--   **Factory vs API**: Verify that creating a project handle via the factory doesn't trigger API calls, but methods like `project.listScreens()` do.
+
+- **Handle Creation**: Verify that `stitch.project('invalid-id')` does not throw (lazy instantiation) but the first call on it fails safely.
+- **Factory vs API**: Verify that creating a project handle via the factory doesn't trigger API calls, but methods like `project.listScreens()` do.
 
 ### 3. Screen Lifecycle (`Screen`)
--   **The Handover**: Verify that properties from `Project.generate()` are correctly populated on the returned `Screen` instances without a second fetch.
--   **Null Safety in Projections**: Test tools or scenarios that return empty arrays or missing optional fields. Verify that the SDK handle handles them as `undefined` or empty arrays rather than crashing on null property access!
+
+- **The Handover**: Verify that properties from `Project.generate()` are correctly populated on the returned `Screen` instances without a second fetch.
+- **Null Safety in Projections**: Test tools or scenarios that return empty arrays or missing optional fields. Verify that the SDK handle handles them as `undefined` or empty arrays rather than crashing on null property access!
 
 ### 4. Design System (`DesignSystem`)
--   **Application**: Create a design system, and apply it to a list of screens. Verify that if the list is empty or invalid, the SDK fails cleanly!
--   **Handles**: Verify that `project.designSystem('ds-id')` correctly receives the `projectId` and injects it into calls like `ds.apply(...)`.
+
+- **Application**: Create a design system, and apply it to a list of screens. Verify that if the list is empty or invalid, the SDK fails cleanly!
+- **Handles**: Verify that `project.designSystem('ds-id')` correctly receives the `projectId` and injects it into calls like `ds.apply(...)`.
 
 ---
 
 ## Tricky Situations Matrix (Standard Functional Edges)
 
-| Scenario | What to try | Expected Behavior |
-| --- | --- | --- |
-| **Stale Handles** | Create a screen, delete the project, then try to edit the screen handle. | Clean API error indicating resource not found, wrapped in `StitchError`. |
-| **Empty Prompts** | Call `project.generate('')` or with only whitespace. | Safe rejection or clear API error, no crash in codegen. |
+| Scenario                | What to try                                                                                                                              | Expected Behavior                                                                                                                                    |
+| ----------------------- | ---------------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------- |
+| **Stale Handles**       | Create a screen, delete the project, then try to edit the screen handle.                                                                 | Clean API error indicating resource not found, wrapped in `StitchError`.                                                                             |
+| **Empty Prompts**       | Call `project.generate('')` or with only whitespace.                                                                                     | Safe rejection or clear API error, no crash in codegen.                                                                                              |
 | **Projections on null** | Force an API call that returns a response without the expected projection field (if you can simulate or find such a tool fallback case). | The SDK should use optional chaining (e.g., `raw?.prop`) and return `undefined` rather than throwing `TypeError: cannot read property of undefined`. |
-| **Massive arrays** | Pass hundreds of screen IDs to `ds.apply()`. | Check if it hits payload limits gracefully or fails with a clear message. |
+| **Massive arrays**      | Pass hundreds of screen IDs to `ds.apply()`.                                                                                             | Check if it hits payload limits gracefully or fails with a clear message.                                                                            |
 
 ---
 
 ## Diagnostic Hygiene
 
--   Always wrap your test calls in `try/catch`.
--   Log the error and inspect `error.code` or `error.name` to see if it's a `StitchError` or a generic raw error.
--   If an execution throws a raw `TypeError` or "cannot read property of undefined", that is a **HIGH PRIORITY BUG** in the SDK's projection logic!
+- Always wrap your test calls in `try/catch`.
+- Log the error and inspect `error.code` or `error.name` to see if it's a `StitchError` or a generic raw error.
+- If an execution throws a raw `TypeError` or "cannot read property of undefined", that is a **HIGH PRIORITY BUG** in the SDK's projection logic!
 
 ---
 
@@ -74,7 +79,7 @@ async function bash() {
   try {
     // 1. Create a fresh project
     project = await stitch.createProject({
-      displayName: `Bug Bash ${new Date().toISOString()}`
+      displayName: `Bug Bash ${new Date().toISOString()}`,
     });
     console.log(`✓ Created Project: ${project.id}`);
 
@@ -89,7 +94,7 @@ async function bash() {
     // 3. Create a design system
     const ds = await project.createDesignSystem({
       name: "Bash Style",
-      variables: { primaryColor: "#ff0000" }
+      variables: { primaryColor: "#ff0000" },
     });
     console.log(`✓ Created Design System: ${ds.id}`);
 
@@ -104,15 +109,14 @@ async function bash() {
     } catch (e) {
       console.log("✗ Did applying to empty list fail? Inspect error.");
     }
-
   } catch (error) {
     console.error("💥 Bash failed with error:", error);
   } finally {
     // 6. Cleanup
     if (project) {
       console.log(`🧹 Cleaning up project ${project.id}...`);
       // Assuming we have a deleteProject binding or we just leave it if not available
-      // await project.delete(); 
+      // await project.delete();
     }
   }
 }

diff --git a/.agents/skills/stitch-sdk-development/SKILL.md b/.agents/skills/stitch-sdk-development/SKILL.md
@@ -51,13 +51,17 @@ When the Stitch MCP server adds a new tool:
 `domain-map.json` expresses two things:
 
 **Classes**: What domain objects exist and how they're constructed.
+
 ```json
 {
   "Screen": {
     "constructorParams": ["projectId", "screenId"],
     "fieldMapping": {
       "projectId": { "from": "projectId" },
-      "screenId": { "from": "id", "fallback": { "field": "name", "splitOn": "/screens/" } }
+      "screenId": {
+        "from": "id",
+        "fallback": { "field": "name", "splitOn": "/screens/" }
+      }
     },
     "parentField": "projectId",
     "idField": "screenId"
@@ -66,6 +70,7 @@ When the Stitch MCP server adds a new tool:
 ```
 
 **Bindings**: How MCP tools map to class methods.
+
 ```json
 {
   "tool": "generate_screen_from_text",
@@ -74,7 +79,10 @@ When the Stitch MCP server adds a new tool:
   "args": {
     "projectId": { "from": "self" },
     "prompt": { "from": "param" },
-    "name": { "from": "computed", "template": "projects/{projectId}/screens/{screenId}" }
+    "name": {
+      "from": "computed",
+      "template": "projects/{projectId}/screens/{screenId}"
+    }
   },
   "returns": {
     "class": "Screen",
@@ -92,9 +100,13 @@ When the Stitch MCP server adds a new tool:
 **Response projections**: Structured `ProjectionStep[]` arrays validated against `outputSchema`. Use `index` for single items, `each` for arrays. Empty `[]` = direct return.
 
 **Cache**: Methods can specify a `cache` with a structured `projection` to check `this.data` before calling the API:
+
 ```json
 {
-  "cache": { "projection": [{ "prop": "htmlCode" }, { "prop": "downloadUrl" }], "description": "Use cached download URL from generation response" }
+  "cache": {
+    "projection": [{ "prop": "htmlCode" }, { "prop": "downloadUrl" }],
+    "description": "Use cached download URL from generation response"
+  }
 }
 ```
 
@@ -113,7 +125,7 @@ For AI agents and orchestration scripts. Raw tool pipe. The agent receives tool
 The `stitch` singleton exposes both domain methods and tool methods via a `Proxy`. No instantiation needed — it lazily creates a `StitchToolClient` from env vars on first access.
 
 ```typescript
-import { stitch } from '@google/stitch-sdk';
+import { stitch } from "@google/stitch-sdk";
 
 // Discover available tools
 const { tools } = await stitch.listTools();
@@ -135,9 +147,9 @@ The singleton reads `STITCH_API_KEY` (or `STITCH_ACCESS_TOKEN` + `GOOGLE_CLOUD_P
 For explicit control (multiple clients, custom config, testing), instantiate `StitchToolClient` directly:
 
 ```typescript
-import { StitchToolClient } from '@google/stitch-sdk';
+import { StitchToolClient } from "@google/stitch-sdk";
 
-const client = new StitchToolClient({ apiKey: 'my-key' });
+const client = new StitchToolClient({ apiKey: "my-key" });
 const tools = await client.listTools();
 const result = await client.callTool("create_project", { title: "My App" });
 await client.close();
@@ -152,13 +164,13 @@ await client.close();
 For agents built on the [Vercel AI SDK](https://sdk.vercel.ai/). Transforms MCP tool schemas into AI SDK-compatible tool definitions, enabling plug-and-play with `generateText()`.
 
 ```typescript
-import { stitchTools } from '@google/stitch-sdk/ai';
-import { generateText } from 'ai';
-import { google } from '@ai-sdk/google';
+import { stitchTools } from "@google/stitch-sdk/ai";
+import { generateText } from "ai";
+import { google } from "@ai-sdk/google";
 
 const result = await generateText({
   model: google("gemini-2.0-flash"),
-  tools: stitchTools(),                          // all tools
+  tools: stitchTools(), // all tools
   // or: stitchTools({ include: ["create_project"] })  // filtered
   prompt: "Create a project called My App",
 });
@@ -170,9 +182,9 @@ const result = await generateText({
 
 ```typescript
 const tool = stitch.toolMap.get("create_project");
-tool.params;                              // ToolParam[] — flat, pre-parsed
-tool.params.filter(p => p.required);      // required params only
-tool.inputSchema;                         // raw ToolInputSchema still available
+tool.params; // ToolParam[] — flat, pre-parsed
+tool.params.filter((p) => p.required); // required params only
+tool.inputSchema; // raw ToolInputSchema still available
 ```
 
 The raw `toolDefinitions` array and standalone `toolMap` are also exported from the main entry point.
@@ -228,8 +240,16 @@ The membrane is declared in `domain-map.json` via `sideEffects` on any class wit
   "Project": {
     "extensionPath": "../../src/project-ext.js",
     "sideEffects": [
-      { "method": "uploadImage", "reason": "private_rest", "specPath": "src/spec/upload.ts" },
-      { "method": "downloadAssets", "reason": "filesystem_io", "specPath": "src/spec/download.ts" }
+      {
+        "method": "uploadImage",
+        "reason": "private_rest",
+        "specPath": "src/spec/upload.ts"
+      },
+      {
+        "method": "downloadAssets",
+        "reason": "filesystem_io",
+        "specPath": "src/spec/download.ts"
+      }
     ]
   }
 }
@@ -238,6 +258,7 @@ The membrane is declared in `domain-map.json` via `sideEffects` on any class wit
 Valid `reason` values: `filesystem_io`, `binary_data`, `private_rest`, `complex_orchestration`.
 
 The generator validates at Stage 3:
+
 1. No `sideEffect.method` collides with a generated binding method name
 2. Each `specPath` points to an existing file
 
@@ -257,13 +278,13 @@ Follow the Spec → Handler → Extension pattern:
 
 #### Rules
 
-| Rule | Rationale |
-|---|---|
-| Extension methods must NOT override generated methods | Prevents silent shadowing |
-| Extension methods must delegate to a Handler | Prevents inline business logic |
-| Handlers must implement a Spec interface | Typed service contract |
-| Handlers must return Result, never throw | Consistent error surface |
-| Extensions must NOT import from `singleton.ts` | Prevents circular dependencies |
+| Rule                                                  | Rationale                      |
+| ----------------------------------------------------- | ------------------------------ |
+| Extension methods must NOT override generated methods | Prevents silent shadowing      |
+| Extension methods must delegate to a Handler          | Prevents inline business logic |
+| Handlers must implement a Spec interface              | Typed service contract         |
+| Handlers must return Result, never throw              | Consistent error surface       |
+| Extensions must NOT import from `singleton.ts`        | Prevents circular dependencies |
 
 ---
 
@@ -323,7 +344,8 @@ Do not rely on cached descriptions of files or directory trees. Read the source.
 ## Import Convention
 
 Use `.js` extensions for ESM compatibility:
+
 ```typescript
-import { StitchError } from '../../src/spec/errors.js';  // ✓
-import { StitchError } from '../../src/spec/errors';     // ✗
+import { StitchError } from "../../src/spec/errors.js"; // ✓
+import { StitchError } from "../../src/spec/errors"; // ✗
 ```