[pull] main from microsoft:main

.github/skills/add-policy/SKILL.md

@@ -230,3 +230,7 @@ See `chat.tools.global.autoApprove` and `chat.useHooks` in `src/vs/workbench/con
 ## Examples
 
 Search the codebase for `policy:` to find all the examples of different policy configurations.
+
+## Learnings
+
+* Never hand-edit `build/lib/policies/policyData.jsonc` (its header explicitly forbids it). If `npm run export-policy-data` is failing, fix the script — don't patch the JSON. Common cause: running it in the wrong working directory (e.g. main repo instead of a worktree), which silently exports the wrong source tree.

.github/workflows/pr-linux-cli-test.yml

@@ -11,7 +11,7 @@ on:
 jobs:
   linux-cli-test:
     name: ${{ inputs.job_name }}
-    runs-on: [ self-hosted, 1ES.Pool=1es-vscode-oss-ubuntu-22.04-x64, "JobId=linux-cli-test-${{ inputs.job_name }}-${{ github.run_id }}-${{ github.run_number }}-${{ github.run_attempt }}" ]
+    runs-on: ubuntu-22.04
     env:
       RUSTUP_TOOLCHAIN: ${{ inputs.rustup_toolchain }}
     steps:

.github/workflows/pr-node-modules.yml

@@ -10,7 +10,7 @@ permissions: {}
 jobs:
   compile:
     name: Compile
-    runs-on: [ self-hosted, 1ES.Pool=1es-vscode-oss-ubuntu-22.04-x64, "JobId=compile-${{ github.run_id }}-${{ github.run_number }}-${{ github.run_attempt }}" ]
+    runs-on: ubuntu-22.04
     steps:
       - name: Checkout microsoft/vscode
         uses: actions/checkout@v6
@@ -86,7 +86,7 @@ jobs:
 
   linux:
     name: Linux
-    runs-on: [ self-hosted, 1ES.Pool=1es-vscode-oss-ubuntu-22.04-x64, "JobId=linux-${{ github.run_id }}-${{ github.run_number }}-${{ github.run_attempt }}" ]
+    runs-on: ubuntu-22.04
     env:
       NPM_ARCH: x64
       VSCODE_ARCH: x64
@@ -219,7 +219,7 @@ jobs:
 
   windows:
     name: Windows
-    runs-on: [ self-hosted, 1ES.Pool=1es-vscode-oss-windows-2022-x64, "JobId=windows-${{ github.run_id }}-${{ github.run_number }}-${{ github.run_attempt }}" ]
+    runs-on: [ self-hosted, 1ES.Pool=1es-vscode-oss-windows-2022-x64 ]
     env:
       NPM_ARCH: x64
       VSCODE_ARCH: x64

.github/workflows/pr-win32-test.yml

@@ -17,7 +17,7 @@ on:
 jobs:
   windows-test:
     name: ${{ inputs.job_name }}
-    runs-on: [ self-hosted, 1ES.Pool=1es-vscode-oss-windows-2022-x64, "JobId=windows-test-${{ inputs.job_name }}-${{ github.run_id }}-${{ github.run_number }}-${{ github.run_attempt }}" ]
+    runs-on: windows-2022
     env:
       ARTIFACT_NAME: ${{ (inputs.electron_tests && 'electron') || (inputs.browser_tests && 'browser') || (inputs.remote_tests && 'remote') || 'unknown' }}
       NPM_ARCH: x64

.github/workflows/pr.yml

@@ -19,7 +19,7 @@ env:
 jobs:
   compile:
     name: Compile & Hygiene
-    runs-on: [ self-hosted, 1ES.Pool=1es-vscode-oss-ubuntu-22.04-x64, "JobId=compile-${{ github.run_id }}-${{ github.run_number }}-${{ github.run_attempt }}" ]
+    runs-on: ubuntu-22.04
     steps:
       - name: Checkout microsoft/vscode
         uses: actions/checkout@v6
@@ -159,7 +159,7 @@ jobs:
 
   copilot-check-test-cache:
     name: Copilot - Check Test Cache
-    runs-on: [ self-hosted, 1ES.Pool=1es-vscode-oss-ubuntu-22.04-x64, "JobId=copilot-check-test-cache-${{ github.run_id }}-${{ github.run_number }}-${{ github.run_attempt }}" ]
+    runs-on: ubuntu-22.04
     permissions:
       contents: read
       pull-requests: read
@@ -205,7 +205,7 @@ jobs:
 
   copilot-check-telemetry:
     name: Copilot - Check Telemetry
-    runs-on: [ self-hosted, 1ES.Pool=1es-vscode-oss-ubuntu-22.04-x64, "JobId=copilot-check-telemetry-${{ github.run_id }}-${{ github.run_number }}-${{ github.run_attempt }}" ]
+    runs-on: ubuntu-22.04
     permissions:
       contents: read
     steps:
@@ -224,7 +224,7 @@ jobs:
 
   copilot-linux-tests:
     name: Copilot - Test (Linux)
-    runs-on: [ self-hosted, 1ES.Pool=1es-vscode-oss-ubuntu-22.04-x64, "JobId=copilot-linux-tests-${{ github.run_id }}-${{ github.run_number }}-${{ github.run_attempt }}" ]
+    runs-on: ubuntu-22.04
     permissions:
       contents: read
     steps:
@@ -329,7 +329,7 @@ jobs:
 
   copilot-windows-tests:
     name: Copilot - Test (Windows)
-    runs-on: [ self-hosted, 1ES.Pool=1es-vscode-oss-windows-2022-x64, "JobId=copilot-windows-tests-${{ github.run_id }}-${{ github.run_number }}-${{ github.run_attempt }}" ]
+    runs-on: windows-2022
     permissions:
       contents: read
     steps:

build/lib/i18n.resources.json

@@ -434,6 +434,10 @@
 			"name": "vs/workbench/services/language",
 			"project": "vscode-workbench"
 		},
+		{
+			"name": "vs/workbench/services/policies",
+			"project": "vscode-workbench"
+		},
 		{
 			"name": "vs/workbench/services/progress",
 			"project": "vscode-workbench"

build/lib/policies/policyData.jsonc

@@ -68,6 +68,21 @@
             "default": "",
             "included": false
         },
+        {
+            "key": "chat.approvedAccountOrganizations",
+            "name": "ChatApprovedAccountOrganizations",
+            "category": "InteractiveSession",
+            "minimumVersion": "1.118",
+            "localization": {
+                "description": {
+                    "key": "chat.approvedAccountOrganizations.policy.description",
+                    "value": "Setting this policy to a non-empty list activates the Approved Account gate: all AI features are disabled until the user signs into a GitHub account whose organizations intersect this list AND the account-side policy data has resolved. Comparison is case-insensitive. Use '*' as a wildcard to accept any signed-in GitHub or GHE account (use this for GHE deployments where the organization list is not surfaced)."
+                }
+            },
+            "type": "array",
+            "default": [],
+            "included": false
+        },
         {
             "key": "extensions.allowed",
             "name": "AllowedExtensions",

extensions/copilot/package.json

@@ -4385,6 +4385,16 @@
 							"onExp"
 						]
 					},
+					"github.copilot.chat.executionSubagent.useAgenticProxy": {
+						"type": "boolean",
+						"default": false,
+						"markdownDescription": "%github.copilot.config.executionSubagent.useAgenticProxy%",
+						"tags": [
+							"advanced",
+							"experimental",
+							"onExp"
+						]
+					},
 					"github.copilot.chat.executionSubagent.toolCallLimit": {
 						"type": "number",
 						"default": 10,
@@ -4574,7 +4584,7 @@
 					},
 					"github.copilot.chat.cli.showExternalSessions": {
 						"type": "boolean",
-						"default": false,
+						"default": true,
 						"markdownDescription": "%github.copilot.config.cli.showExternalSessions%",
 						"tags": [
 							"advanced"
@@ -4604,6 +4614,14 @@
 							"advanced"
 						]
 					},
+					"github.copilot.chat.cli.lazyLoadSessionItem.enabled": {
+						"type": "boolean",
+						"default": false,
+						"markdownDescription": "%github.copilot.config.cli.lazyLoadSessionItem.enabled%",
+						"tags": [
+							"advanced"
+						]
+					},
 					"github.copilot.chat.cli.aiGenerateBranchNames.enabled": {
 						"type": "boolean",
 						"default": true,
@@ -4639,7 +4657,7 @@
 					},
 					"github.copilot.chat.cli.sessionController.enabled": {
 						"type": "boolean",
-						"default": false,
+						"default": true,
 						"markdownDescription": "%github.copilot.config.cli.sessionController.enabled%",
 						"tags": [
 							"advanced"

extensions/copilot/package.nls.json

@@ -411,6 +411,7 @@
 	"github.copilot.config.cli.planExitMode.enabled": "Enable Plan Mode exit handling in Copilot CLI.",
 	"github.copilot.config.cli.autoModel.enabled": "Enable the Auto model option in Copilot CLI, which automatically selects the best model for each request. Requires VS Code reload.",
 	"github.copilot.config.cli.planCommand.enabled": "Enable the /plan command in Copilot CLI to create implementation plans before coding.",
+	"github.copilot.config.cli.lazyLoadSessionItem.enabled": "Enable lazy loading of session items in Copilot CLI. Requires VS Code reload.",
 	"github.copilot.config.cli.aiGenerateBranchNames.enabled": "Enable AI-generated branch names in Copilot CLI.",
 	"github.copilot.config.cli.isolationOption.enabled": "Enable the isolation mode option for Copilot CLI. When enabled, users can choose between Worktree and Workspace modes.",
 	"github.copilot.config.cli.autoCommit.enabled": "Enable automatic commit for Copilot CLI. When enabled, changes made by Copilot CLI will be automatically committed to the repository at the end of each turn.",
@@ -492,7 +493,8 @@
 	"copilot.tools.executionSubagent.name": "Execution Subagent",
 	"copilot.tools.executionSubagent.description": "Launch an execution-focused subagent that runs one or more terminal commands to accomplish a task. It is designed to select an efficient summary of the terminal outputs to return to the main agent context.",
 	"github.copilot.config.executionSubagent.enabled": "Enable the Execution Subagent tool in Copilot Chat. The Execution Subagent is designed to run terminal commands to accomplish an execution-based task.",
-	"github.copilot.config.executionSubagent.model": "The model to use for the Execution Subagent tool in Copilot Chat. Leave empty to use the default model.",
+	"github.copilot.config.executionSubagent.useAgenticProxy": "Use the agentic proxy endpoint for the execution subagent.",
+	"github.copilot.config.executionSubagent.model": "The model to use for the Execution Subagent tool in Copilot Chat. When useAgenticProxy is enabled, defaults to 'exec-subagent-router-a'. Otherwise defaults to the main agent model.",
 	"github.copilot.config.executionSubagent.toolCallLimit": "Maximum number of tool calls the Execution Subagent can make during execution.",
 	"github.copilot.session.providerDescription.claude": "Delegate tasks to the Claude SDK running locally on your machine. The agent iterates via chat and works asynchronously to implement changes.",
 	"github.copilot.session.providerDescription.cloud": "Delegate tasks to the GitHub Copilot coding agent. The agent iterates via chat and works asynchronously in the cloud to implement changes and pull requests as needed.",

extensions/copilot/src/extension/chatSessions/claude/AGENTS.md

@@ -214,6 +214,8 @@ In multi-root and empty workspaces, a folder picker option appears in the chat s
 ### Key Files
 
 - **`common/claudeFolderInfo.ts`**: `ClaudeFolderInfo` interface
+- **`../../chatSessions/common/claudeWorkspaceFolderService.ts`**: `IClaudeWorkspaceFolderService` interface — computes git diff changes for session items
+- **`../../chatSessions/vscode-node/claudeWorkspaceFolderServiceImpl.ts`**: Implementation — diffs the session's branch against its base branch, caches results, and maps changes to `ChatSessionChangedFile[]` for display in the Sessions view
 - **`../../chatSessions/vscode-node/claudeChatSessionContentProvider.ts`**: Folder resolution, picker options, and handler integration
 - **`../../chatSessions/vscode-node/folderRepositoryManagerImpl.ts`**: `FolderRepositoryManager` (abstract base) with `ClaudeFolderRepositoryManager` subclass — the Claude subclass does not depend on `ICopilotCLISessionService` (CopilotCLI has its own subclass `CopilotCLIFolderRepositoryManager`)
 - **`node/claudeCodeAgent.ts`**: Consumes `ClaudeFolderInfo` in `ClaudeCodeSession._startSession()`

extensions/copilot/src/extension/chatSessions/claude/CLAUDE_SESSION_USER_GUIDE.md

@@ -225,6 +225,7 @@ Each session in the list displays:
 | **Blue dot** | Indicates an unread or recently active session |
 | **Status icon** | Shows whether the session is completed, in progress, needs input, or failed |
 | **Folder badge** | In multi-root or empty workspaces, shows which folder the session ran in |
+| **Change stats** | Shows lines added and removed (e.g., `+584 -17`) — a quick summary of the session's code impact, computed by diffing the session's branch against its base branch |
 
 Sessions are sorted by recency — the most recent session appears at the top. In the dedicated sidebar, they're also grouped by time period.
 

extensions/copilot/src/extension/chatSessions/claude/common/claudeSessionStateService.ts

@@ -3,7 +3,7 @@
  *  Licensed under the MIT License. See License.txt in the project root for license information.
  *--------------------------------------------------------------------------------------------*/
 
-import { PermissionMode } from '@anthropic-ai/claude-agent-sdk';
+import { EffortLevel, PermissionMode } from '@anthropic-ai/claude-agent-sdk';
 import type * as vscode from 'vscode';
 import { CapturingToken } from '../../../../platform/requestLogger/common/capturingToken';
 import { createServiceIdentifier } from '../../../../util/common/services';
@@ -22,7 +22,7 @@ export interface SessionState {
 	capturingToken: CapturingToken | undefined;
 	folderInfo: ClaudeFolderInfo | undefined;
 	usageHandler: UsageHandler | undefined;
-	reasoningEffort: string | undefined;
+	reasoningEffort: EffortLevel | undefined;
 }
 
 /**
@@ -96,12 +96,12 @@ export interface IClaudeSessionStateService {
 	/**
 	 * Gets the reasoning effort for a session (user's per-request selection from the model picker).
 	 */
-	getReasoningEffortForSession(sessionId: string): string | undefined;
+	getReasoningEffortForSession(sessionId: string): EffortLevel | undefined;
 
 	/**
 	 * Sets the reasoning effort for a session.
 	 */
-	setReasoningEffortForSession(sessionId: string, effort: string | undefined): void;
+	setReasoningEffortForSession(sessionId: string, effort: EffortLevel | undefined): void;
 }
 
 export const IClaudeSessionStateService = createServiceIdentifier<IClaudeSessionStateService>('IClaudeSessionStateService');

extensions/copilot/src/extension/chatSessions/claude/node/claudeCodeAgent.ts

@@ -3,7 +3,7 @@
  *  Licensed under the MIT License. See License.txt in the project root for license information.
  *--------------------------------------------------------------------------------------------*/
 
-import { McpServerConfig, Options, PermissionMode, Query, SDKUserMessage } from '@anthropic-ai/claude-agent-sdk';
+import { EffortLevel, McpServerConfig, Options, PermissionMode, Query, SDKUserMessage } from '@anthropic-ai/claude-agent-sdk';
 import Anthropic from '@anthropic-ai/sdk';
 import * as l10n from '@vscode/l10n';
 import type * as vscode from 'vscode';
@@ -161,6 +161,7 @@ export class ClaudeCodeSession extends Disposable {
 	private _settingsChangeTracker: ClaudeSettingsChangeTracker;
 	private _currentModelId: ParsedClaudeModelId;
 	private _currentPermissionMode: PermissionMode;
+	private _currentEffort: EffortLevel | undefined;
 	private _isResumed: boolean;
 	private _yieldInProgress = false;
 	private _sessionStarting: Promise<void> | undefined;
@@ -337,7 +338,15 @@ export class ClaudeCodeSession extends Disposable {
 		// Do this BEFORE starting a session so the Options are correct from the start
 		const modelId = this.sessionStateService.getModelIdForSession(this.sessionId);
 		const permissionMode = this.sessionStateService.getPermissionModeForSession(this.sessionId);
+		const effortLevel = this.sessionStateService.getReasoningEffortForSession(this.sessionId);
 
+		if (effortLevel !== this._currentEffort) {
+			this._currentEffort = effortLevel;
+			// Effort doesn't have a direct setter on the query generator, so we need to restart the session
+			if (this._queryGenerator) {
+				this._restartSession();
+			}
+		}
 		// Update model and permission mode on active session if they changed
 		if (modelId) {
 			await this._setModel(modelId);
@@ -427,6 +436,7 @@ export class ClaudeCodeSession extends Disposable {
 			// the permission mode ourselves in the options
 			allowDangerouslySkipPermissions: true,
 			abortController: this._abortController,
+			effort: this._currentEffort,
 			executable: process.execPath as 'node', // get it to fork the EH node process
 			// TODO: CAPI does not yet support the WebSearch tool
 			// Once it does, we can re-enable it.
@@ -662,6 +672,7 @@ export class ClaudeCodeSession extends Disposable {
 		this._queryGenerator = undefined;
 		this._abortController = new AbortController();
 		this._currentRequest = undefined;
+		this._currentEffort = undefined;
 	}
 
 	/**

extensions/copilot/src/extension/chatSessions/claude/node/claudeCodeModels.ts

@@ -13,6 +13,7 @@ import { Emitter } from '../../../../util/vs/base/common/event';
 import { Disposable } from '../../../../util/vs/base/common/lifecycle';
 import type { ParsedClaudeModelId } from '../common/claudeModelId';
 import { tryParseClaudeModelId } from './claudeModelId';
+import { EffortLevel } from '@anthropic-ai/claude-agent-sdk';
 
 export const CLAUDE_REASONING_EFFORT_PROPERTY = 'reasoningEffort';
 
@@ -24,7 +25,13 @@ export interface IClaudeCodeModels {
 	 * then to the newest Sonnet, newest Haiku, or any Claude endpoint.
 	 * Returns `undefined` if no Claude endpoint can be found.
 	 */
-	resolveEndpoint(requestedModel: string | undefined, fallbackModelId: ParsedClaudeModelId | undefined): Promise<IChatEndpoint | undefined>;
+	resolveEndpoint(requestedModel: ParsedClaudeModelId | string | undefined, fallbackModelId: ParsedClaudeModelId | undefined): Promise<IChatEndpoint | undefined>;
+
+	/**
+	 * Resolves the reasoning effort level for the given requested model ID and requested reasoning effort.
+	 */
+	resolveReasoningEffort(requestedModel: ParsedClaudeModelId | string | undefined, requestedReasoningEffort: string | undefined): Promise<EffortLevel | undefined>;
+
 	/**
 	 * Registers a LanguageModelChatProvider so that Claude models appear in
 	 * VS Code's built-in model picker for the claude-code session type.
@@ -101,12 +108,32 @@ export class ClaudeCodeModels extends Disposable implements IClaudeCodeModels {
 		});
 	}
 
-	public async resolveEndpoint(requestedModel: string | undefined, fallbackModelId: ParsedClaudeModelId | undefined): Promise<IChatEndpoint | undefined> {
+	public async resolveReasoningEffort(requestedModel: ParsedClaudeModelId | string | undefined, requestedReasoningEffort: string | undefined): Promise<EffortLevel | undefined> {
+		const endpoint = await this.resolveEndpoint(requestedModel, undefined);
+		if (!endpoint || !endpoint.supportsReasoningEffort || endpoint.supportsReasoningEffort.length === 0) {
+			return undefined;
+		}
+		if (requestedReasoningEffort && isEffortLevel(requestedReasoningEffort) && endpoint.supportsReasoningEffort.includes(requestedReasoningEffort)) {
+			return requestedReasoningEffort;
+		}
+		if (endpoint.supportsReasoningEffort.length === 1 && isEffortLevel(endpoint.supportsReasoningEffort[0])) {
+			return endpoint.supportsReasoningEffort[0];
+		}
+		return undefined;
+	}
+
+	public async resolveEndpoint(requestedModel: ParsedClaudeModelId | string | undefined, fallbackModelId: ParsedClaudeModelId | undefined): Promise<IChatEndpoint | undefined> {
 		const endpoints = await this._getEndpoints();
 
 		// 1. Exact match for the requested model
 		if (requestedModel) {
-			const mappedModel = tryParseClaudeModelId(requestedModel)?.toEndpointModelId() ?? requestedModel;
+			let parsedModel: ParsedClaudeModelId | undefined;
+			if (typeof requestedModel === 'string') {
+				parsedModel = tryParseClaudeModelId(requestedModel);
+			} else {
+				parsedModel = requestedModel;
+			}
+			const mappedModel = parsedModel?.toEndpointModelId() ?? requestedModel;
 			const exact = endpoints.find(e => e.family === mappedModel || e.model === mappedModel);
 			if (exact) {
 				return exact;
@@ -163,14 +190,18 @@ export class ClaudeCodeModels extends Disposable implements IClaudeCodeModels {
 	}
 }
 
-const SUPPORTED_EFFORT_LEVELS = ['low', 'medium', 'high'] as const;
+const SUPPORTED_EFFORT_LEVELS: EffortLevel[] = ['low', 'medium', 'high'];
+
+export function isEffortLevel(value: string): value is EffortLevel {
+	return SUPPORTED_EFFORT_LEVELS.includes(value as EffortLevel);
+}
 
 function buildConfigurationSchema(endpoint: IChatEndpoint): vscode.LanguageModelConfigurationSchema | undefined {
 	const effortLevels = endpoint.supportsReasoningEffort?.filter(
 		(level): level is typeof SUPPORTED_EFFORT_LEVELS[number] =>
 			(SUPPORTED_EFFORT_LEVELS as readonly string[]).includes(level)
 	);
-	if (!effortLevels || effortLevels.length <= 1) {
+	if (!effortLevels) {
 		return;
 	}
 

extensions/copilot/src/extension/chatSessions/claude/node/claudeLanguageModelServer.ts

@@ -215,7 +215,10 @@ export class ClaudeLanguageModelServer extends Disposable {
 			}
 
 			const capturingToken = sessionId ? this.sessionStateService.getCapturingTokenForSession(sessionId) : undefined;
-			const reasoningEffort = sessionId ? this.sessionStateService.getReasoningEffortForSession(sessionId) : undefined;
+			const sessionReasoningEffort = sessionId ? this.sessionStateService.getReasoningEffortForSession(sessionId) : undefined;
+			const reasoningEffort = sessionReasoningEffort && selectedEndpoint.supportsReasoningEffort?.includes(sessionReasoningEffort)
+				? sessionReasoningEffort
+				: undefined;
 
 			const doRequest = () => streamingEndpoint.makeChatRequest2({
 				debugName: 'Claude Copilot Proxy',