main.js · part 130

60 text occurrences from desktop/Cursor.app/Contents/Resources/app/extensions/cursor-always-local/dist/main.js, part 130. Every entry preserves the shipped literal and its saved verdict or selection reason.

File contents and all parts · All files

.pick() cannot be used on object schemas containing refinements

Unrecognized key: "${n}"

.omit() cannot be used on object schemas containing refinements

Unrecognized key: "${t}"

.partial() cannot be used on object schemas containing refinements

Unrecognized key: "${e}"

Unrecognized key: "${e}"

Key ${n} not found in enum

Key ${e} not found in enum

This schema contains multiple valid literal values. Use `.values` instead.

Install Cursor CLI?

curl https://cursor.com/install -fsS | bash

Invalid value for number array: ${t}

Each mcp_reconnect_config override requires identifier and/or hostname.

Expected exactly 3 icon actions, got ${n}

Expected exactly one "${t}" action, got ${n}

Unknown FTUX first-action job role key: ${n}

Rate limited by model provider, retrying…

Use /in-cloud for cloud subagents

Authenticate this MCP server so its tools can be used. Call this tool through your MCP tool-calling interface when STATUS.md indicates this server needs authentication.

The MCP server errored. If this server is important for completing the task, concisely inform the user and ask them to check the MCP status in Cursor's Customize page > MCPs; otherwise continue with a different approach.

The MCP server needs authentication. Authenticate it by calling the `{authToolName}` tool for server "{serverIdentifier}" through your MCP tool-calling interface using an empty arguments object. If this server is important for completing the task, authenticate it first; otherwise continue with a different approach.

Issue reproduced, please proceed

The issue has been fixed. Please clean up the instrumentation.

You are analyzing conversation transcripts to identify repeated user behaviors.

## Transcript Location
Transcripts are stored at: {{agentTranscriptsPath}}

## Transcript Format
Each '.txt' file is a human-readable conversation transcript with this structure:
- 'user:' sections contain user messages (often wrapped in '<user_query>' tags)
- 'assistant:' sections contain assistant responses
- '[Tool call]' blocks show which tools were invoked
- '[Tool result]' blocks show tool outputs

Files can be large. Focus on extracting the '<user_query>' sections which contain the actual user requests. Do NOT try to read the entire file contents because it will pollute your context.

## How to Read Transcripts
1. Use Glob to list files: "{{agentTranscriptsPath}}/*.txt"
2. For each file, extract just the '<user_query>' blocks - these show what users asked for
3. You can use Grep to search for '<user_query>' patterns across files

## Task
Analyze the {{maxChatsToRead}} most recent conversations (by file modification time). Be thorough and do not bias towards recency when analyzing conversations. Ignore trivial conversations and conversations where nothing concrete happened.

Your goal is to understand how the user interacts with the agent. Focus especially on the corrections that the user makes repeatedly. Focus on common terminal commands/workflows the user instructs the agent to do, changes, and other very stable patterns.

## Evidence Standards
When making claims about what the agent SHOULD or SHOULD NOT do, only cite conversations where:
- The user explicitly corrected the agent for doing something wrong
- The user undid or rejected an agent action
- The user gave an explicit instruction ('don\'t do X', 'always do Y')

Do NOT infer 'don\'t do X' from:
- The user asking a question about X (e.g., 'should we add tests?' does not mean 'don\'t add tests proactively')
- The user doing X themselves (doesn't mean the agent shouldn't)
- Absence of the agent doing X

For each claim, ask: 'Is there a conversation where the user pushed back on the agent for doing this?' If not, don't include it as a guideline. For each piece of information you discern, you must cite 4 conversations. For each citation, verify that the conversation actually backs up your claim. Never include direct quotes from user messages; summarize at a high level.

These are the sections to cover. Do not overlap these with existing user and project rules; if there is overlap or conflict always go with the existing rules.

- Developer profile: How does the user interact with the agent. Focus on how much autonomy they like to give the agent vs. how much they would like to oversee the changes the agent is making. What is their workflow for getting tasks done. What kind of tasks do they often work on. Verify across many conversations, and do not extrapolate too hard.
- Frequented areas of the codebase: Parts of the codebase the user primarily works in, and what kinds of tasks correspond to each part of the codebase.
- Important terminal commands: Terminal commands/workflows the user runs repeatedly that are unique to the project/user workflow, and when they should be used. Focus on test, lint, and build commands. Do not include git commands.

Do not write any files in this step. You will be asked in a follow-up message to write the final user profile to disk.

Turn your analysis into a concise 'index.md' markdown file that will be shown to all agents in the future.

Write the file contents to: {{userIntentDirPath}}
(This is a temporary file path and will be atomically moved to {{finalUserIntentDirPath}}.)

Requirements:
- Do not cite user messages or transcripts directly (no quotes)
- Avoid overly specific task details; focus on stable patterns and preferences
- Keep it reasonably short
- For each guideline/claim, include 4 conversation citations and verify that each citation supports the claim

Autopilot PR

Fix CI, conflicts, and comments with full self driving

Autopilot PR runtime

Run in Cloud

Run locally in a Worktree

Debug CI Failure

Debug {count} CI Failures

Debugging CI

Debugging CI

Fix CI with FSD

Fix {count} CI Failures with FSD

Fixing CI with FSD

Fixing CI with FSD

Grok Bot can do this for you

An AI teammate that works in your tools and comes back with finished work.

Open Grok Bot

You are answering a single ephemeral question about the user's current work. Use the provided conversation as context and respond directly in one answer. Do not ask follow-up questions, do not request mode switches, and do not mention tool limitations unless absolutely necessary.

If the conversation does not contain enough to answer, say so briefly.

A Cursor Canvas is a live React app that the user can open beside the chat. You MUST use a canvas when the agent produces a standalone analytical artifact — quantitative analyses, billing investigations, security audits, architecture reviews, data-heavy content, timelines, charts, tables, interactive explorations, repeatable tools, or any response that benefits from visual layout. Especially prefer a canvas when presenting results from MCP tools (Datadog, Databricks, Linear, Sentry, Slack, etc.) where the data is the deliverable — render it in a rich canvas rather than dumping it into a markdown table or code block. If you catch yourself about to write a markdown table, stop and use a canvas instead. You MUST also read this skill whenever you create, edit, or debug any .canvas.tsx file.

The canvas at `{canvasPath}` has the following error:

Check if the canvas SDK has changed since this canvas was created.

Update the canvas to use the latest SDK components according to the supplied documentation in the canvas skill.

Invalid value for enum: ${e}, expected one of: ${t.join(", ")}

Invalid value for boolean: ${t}

Invalid value for string: ${t}

Invalid value for number: ${t}

Cursor-always-local debug logs

Cursor Tab

SpoofedIdProvider not initialized

Using spoofed token

Cursor Structured Logs

no last seen

[TelemChunkReassembler] Purging stale request ${r.slice(0,50)}... (had ${e.chunks.size} chunks, ${n}, age: ${((t-e.createdAt)/1e3).toFixed(1)}s)

Missing chunk ${n} for reassembly