Add support for continue, stopReason and systemMessage for tool hooks

[pwang347]

No description provided.

[Copilot]

Pull request overview

This PR extends chat tool hook handling so hook commands can signal “stop processing” and surface hook warnings/system messages in the chat UI, and updates hook output validation and tests accordingly.

Changes:

• Add continue to common hook output validation and map systemMessage/warnings into hook results.
• Introduce hook-progress events and a workbench contribution to append hook progress parts into the chat transcript.
• Update tool invocation flow to stop the agent loop when hooks indicate stopping, and expand unit tests for the new semantics.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
src/vs/workbench/contrib/chat/test/common/hooksExecutionService.test.ts	Updates expectations around proxy errors/warnings, continue:false, and stop semantics.
src/vs/workbench/contrib/chat/common/hooks/hooksTypes.ts	Adds continue?: boolean to the common hook output validator.
src/vs/workbench/contrib/chat/common/hooks/hooksExecutionService.ts	Adds hook-progress eventing and stop semantics; changes error/warning mapping and aggregation behavior.
src/vs/workbench/contrib/chat/browser/tools/languageModelToolsService.ts	Attempts to stop/cancel tool invocation when hook result includes stopReason.
src/vs/workbench/contrib/chat/browser/chat.contribution.ts	Registers a workbench contribution to show hook progress in the chat UI.

Comments suppressed due to low confidence (2)

src/vs/workbench/contrib/chat/browser/tools/languageModelToolsService.ts:515

• Same issue as preToolUse: executePostToolUseHook throws HookAbortError on stopReason, so this check won’t run. Either handle the thrown error here, or have the hook execution service return stopReason instead of throwing so this logic can work.

			return;
		}

		const hookInput: IPostToolUseCallerInput = {
			toolName: dto.toolId,
			toolInput: dto.parameters,
			getToolResponseText: () => toolContentToA11yString(toolResult.content),
			toolCallId: dto.callId,

src/vs/workbench/contrib/chat/common/hooks/hooksExecutionService.ts:301

• Setting stopReason = '' to signal “stop without a reason” won’t actually stop remaining hooks today because executeHook checks if (result.stopReason) (truthy) before breaking. If you keep the empty-string sentinel, the stopping logic needs to treat stopReason !== undefined as “stop”, not truthiness.

				// stopReason takes precedence if both are set
				let stopReason = commonFields.stopReason;
				if (commonFields.continue === false && !stopReason) {
					stopReason = '';  // Empty string signals stop without a specific reason
				}