main.js · part 216

60 text occurrences from desktop/Cursor.app/Contents/Resources/app/extensions/cursor-agent-host/dist/main.js, part 216. Every entry preserves the shipped literal and its saved verdict or selection reason.

File contents and all parts · All files

The 'ask' permission for subagentStart hooks is not yet implemented. Use 'allow' or 'deny' instead.

Subagent blocked - 'ask' permission not implemented

subagentStop hook firing

subagentStop hook returned unexpected response

subagentStop hook completed

Step blob not found when extracting result

`[Name](${e})`. Don't use a generic label such as `[agent]`, `[worker]`, or `[subagent]`. For cloud subagents, when the agent has edited code, link to `[Review](${e}#changes)`, or, if you know the exact added and deleted line counts, `[Review +A −D](${e}#changes)`, replacing A and D with those counts. Never write A or D literally. Use `[Try Live](${e}#desktop)` only when the agent used computer use

Agent ID: ${s.agentId}${r?` (to link to this agent/subagent in user-facing text use ${n(s.agentId)})`:""}

 (to link to this agent/subagent in user-facing text use ${n(s.agentId)})

Agent ID: ${s.agentId} (can be used with the `resume` parameter to send a follow-up after it completes${r?`, or to link to this agent/subagent in user-facing text with ${n(s.agentId)}`:""})

, or to link to this agent/subagent in user-facing text with ${n(s.agentId)}

${s.resultSuffix}

The user manually backgrounded the subagent. It is still running; you can continue with other work.

Subagent is running in the background.

${n} If needed, you can monitor its output by tailing the transcript at: ${s}. When you end your turn, you will be automatically sent the subagent's final response upon its completion, so do not wait for it - either end your turn or work on something else.
Do NOT mention the transcript path to the user. Do NOT try to predict the subagent's response before it replies.

${n} When you end your turn, you will be automatically sent the subagent's final response upon its completion, so do not wait for it - either end your turn or work on something else.

${n} You can monitor its output by tailing the transcript at: ${s}. Do not mention the transcript path to the user.

No output

This is the output of the subagent:

${a}${o}${i}

Error: ${e.result.value.error}

Unknown error

Unhandled result case: ${t}

Provider error:

Not found error:

API Error:

${e}. provider_status=${r}

Unhandled interaction query type: ${t.query.case}

${e}
${Ne.f9i}

Invalid arguments:
resume: "${bB}" is not supported. Use a valid requestId or do not specify the resume parameter.

Invalid arguments:
machine replaces ${t.join(", ")}; specify machine alone.

Invalid arguments:
cloud_base_branch may only be specified when environment equals cloud.

Invalid arguments:
cloud_requested_environment_build_id may only be specified when environment equals cloud.

Run on this machine, sharing its checkout and branch. The default.

Worker to run on, from cursor-cloud-list-self-hosted-workers. Only your own machines can be targeted this way; use self_hosted_pool for a team pool worker. Check that tool's sharedAssignmentAllowed first: a shared worker runs this subagent alongside others, otherwise the subagent waits for the worker to free up.

Run on one specific self-hosted worker of your own. The subagent uses that machine's existing checkout and branch, so it cannot be given a base branch.

Pool to draw a worker from. Defaults to the team's default pool.

Key/value labels a candidate worker must all match.

Run on any free worker in a self-hosted pool. Pool workers are claimed exclusively, so the subagent queues until one is free.

, or in the most recent <user_info_catalog_update> block on a later user turn

The available model slugs are listed in <available_subagent_models> in the initial user-info message${n}. 

The available model slugs are listed in this tool's description. 

An invalid slug fails the call with the allowed list — pick a listed slug or omit this parameter; never retry an invalid slug unchanged. 

Optional model slug for this agent. If provided, it must resolve to one of the available model slugs. ${s}${o}If omitted, the subagent uses the same model as the parent agent. Do not pass if resume field is set (prior model will be used). Use "inherit" unless the user explicitly requested another listed model.

Optional model slug for this agent. If provided, it must resolve to one of the available model slugs. ${s}${o}If omitted, the subagent uses the same model as the parent agent. Do not pass if resume field is set (prior model will be used). Only choose an explicit model when the user directly requests it.

prompt is required

The task for the agent to perform

Optional agent ID to resume from. If provided, sends a follow-up message to the agent after it has completed. Requests to a currently running asynchronous agent 

fail unless `interrupt` is true; set `interrupt` to true only when you intend to interrupt the running agent.

fail; wait for completion before resuming.

Optional agent ID to resume from. If provided, the agent will continue from the previous execution transcript.

${h} Use "${bB}" to start a new agent with your own entire conversation history as a starting point (aka 'self-fork').

If true and `resume` targets a running async agent, interrupt the current run and send this prompt immediately. Only use when the user explicitly asks to interrupt or change what the running agent is doing.

Subagent type to use for this task. Must be one of: ${a.join(", ")}.

Subagent type to use for this task. Available types are listed in the initial user-info message${n}.

Optional array of file paths to images or videos to attach to the subagent. Files are read and attached to the subagent's context. Use to forward relevant media to any subagent (e.g. pass a user-attached image's saved file path so the subagent sees the actual image rather than a prose description).

Optional execution environment for the subagent. Use "local" (default) for normal local subagents, or "cloud" to run the subagent as a cloud agent (i.e. in its own separate worktree). ONLY set to cloud if the user explicitly requests a cloud subagent. DO NOT set to cloud if user does not request cloud. Cloud subagents will work on their own git branch on their own VM. After subagent completion, follow user instructions on whether to merge that branch into your own branch, check it out, or neither.

 If you mention an agent or subagent in your response, link it with the `[Name](id)` Don't use generic label such as `[agent]`, `[worker]`, or `[subagent]`.

 For cloud subagents, when the agent has edited code, link to `[Review](bc-id#changes)`, or, if you know the exact added and deleted line counts, `[Review +A −D](bc-id#changes)`, replacing A and D with those counts. Never write A or D literally. Use `[Try Live](bc-id#desktop)` only when the agent used computer use.

Base branch for the cloud subagent's branch to start from. Default is current branch. Uses remote version of branch; uncommitted or un-pushed branches will fail. Only specify this parameter if environment equals cloud.

Exact environment build id (e.g. bld-YYYYMMDD-<uuid>) for the cloud subagent's VM to boot from, instead of the environment's latest successful build. Use to test a specific environment build in an isolated cloud subagent. Only specify this parameter if environment equals cloud. The build must belong to the same team and environment; an invalid or inaccessible build fails the subagent.