fix: update chat slash commands to include GitHub Copilot targets and enhance attachment handling

[TylerLeonhardt]

This may not be the right way to handle this... but look, slash commands that are silent like /clear, /fork, /debug, /models are not attachments. They're simply another way to invoke a command simple as that.

This code originally gated those showing up in completions AND actually working (parser changes) on whether or not the chatSession supported PromptAttachments. This refactors that to allow silent commands to go through even if that's not defined.

2nd. This change then caused allllll these permissiony commands to show up for Claude even though that UX doesn't exist for Claude so I have hard coded the target only because the when clause is also hard coded here:

vscode/src/vs/workbench/contrib/chat/browser/actions/chatExecuteActions.ts

Line 480 in 5ce6509

ChatContextKeys.lockedCodingAgentId.isEqualTo(AgentSessionProviders.Background),

With these 2 changes, Claude gets a few more slash commands:

• /clear
• /fork
• /debug
• /models (and Copilot CLI gets this one too since I removed the hardcoded target)

[Copilot]

Pull request overview

This PR refactors how chat slash commands are parsed and suggested so that “silent” (UI/client-side) commands can still work even when the active/locked chat session does not support prompt attachments, and it adjusts targeting so certain permission-related commands don’t appear for unsupported targets (e.g. Claude).

Changes:

• Update ChatRequestParser to allow silent slash commands to parse even when supportsPromptAttachments is false.
• Adjust slash command completions filtering to only require supportsPromptAttachments for commands that aren’t silent.
• Update some slash command targets (and remove /models target restriction) to better control visibility across targets.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 4 comments.

File	Description
src/vs/workbench/contrib/chat/common/requestParser/chatRequestParser.ts	Alters slash command parsing rules for agent/locked-agent contexts, allowing silent commands without prompt attachments.
src/vs/workbench/contrib/chat/browser/widget/input/editor/chatInputCompletions.ts	Updates completion filtering logic so silent commands can still be suggested without prompt attachment support.
src/vs/workbench/contrib/chat/browser/chatSlashCommands.ts	Adjusts targets for permission commands and removes target restriction for /models to change which sessions see these commands.

Comments suppressed due to low confidence (2)

src/vs/workbench/contrib/chat/common/requestParser/chatRequestParser.ts:248

• Spelling/grammar in the new comments: “If there is an agent, we let” reads incomplete (likely needs a colon), and “asume” should be “assume”.

		// If there is no agent, we allow any slash command.
		// If there is an agent, we let
		// * silent ones go through since they are only UI-facing and don't influence chat history
		// * slash commands that support prompt attachments, since those are meant to be used in conjunction with an agent and we can assume the agent can handle them.
		if (!usedAgent || slashCommand?.silent || capabilities?.supportsPromptAttachments) {
			if (slashCommand) {
				// Valid standalone slash command
				return new ChatRequestSlashCommandPart(slashRange, slashEditorRange, slashCommand);
			} else {
				// check for with default agent for this location
				const defaultAgent = this.agentService.getDefaultAgent(location, context?.mode);
				const subCommand = defaultAgent?.slashCommands.find(c => c.name === command);
				if (subCommand) {
					// Valid default agent subcommand
					return new ChatRequestAgentSubcommandPart(slashRange, slashEditorRange, subCommand);
				}
			}

			// if there's no agent or attachments are supported, asume it is a prompt slash command
			const isPromptCommand = this.promptsService.isValidSlashCommandName(command);

src/vs/workbench/contrib/chat/common/requestParser/chatRequestParser.ts:234

• The new gating condition only allows standalone slash commands through when slashCommand.silent is true, but some execute-immediately UI commands (e.g. /clear) are not marked silent and will still be blocked when supportsPromptAttachments is false. If the intent is “client-side/executeImmediately commands should work without prompt attachments”, consider using executeImmediately (or executeImmediately || silent) for this check, or ensure the affected commands are marked silent: true.

		const capabilities = context?.attachmentCapabilities ?? usedAgent?.capabilities ?? context?.attachmentCapabilities;
		const slashCommands = this.slashCommandService.getCommands(location, context?.mode ?? ChatModeKind.Ask);
		const slashCommand = slashCommands.find(c => c.command === command);
		// If there is no agent, we allow any slash command.
		// If there is an agent, we let
		// * silent ones go through since they are only UI-facing and don't influence chat history
		// * slash commands that support prompt attachments, since those are meant to be used in conjunction with an agent and we can assume the agent can handle them.
		if (!usedAgent || slashCommand?.silent || capabilities?.supportsPromptAttachments) {
			if (slashCommand) {