main.js · part 110

60 text occurrences from desktop/Cursor.app/Contents/Resources/app/extensions/cursor-local-agent-runtime/dist/main.js, part 110. Every entry preserves the shipped literal and its saved verdict or selection reason.

File contents and all parts · All files

${e.text}

${n}

Duration of tool call execution in milliseconds

Result status of tool call execution

Approximate number of tokens in raw model-produced tool call arguments (argsText.length / 4)

Tool-call stream chunks whose id collided with an already-handled or in-flight tool call in the same model response

Error writing to tool call stream

Error writing to tool call stream

Tool not found: ${m}. Available tools: ${Object.keys(r).join(", ")}

semantic search usage detected

semantic search usage detected

WritableIterable is closed

Error closing tool call stream

No current stream found

Unknown error

Execution backend unavailable. Tool execution skipped.

Execution backend unavailable. Tool execution skipped.

Unknown tool error

Pending tool contract mismatch for ${e.descriptor.toolName}: expected effective tool ${e.contract.effectiveToolName}, but none resolved

Pending tool contract mismatch for ${t}: expected ${o}, got ${n.toolIdentifier}

[SUGGESTION MODE: Suggest what user might naturally type next into Cursor.]

FIRST: Look at user's recent messages and original request.

Your job is to predict what THEY would type - not what you think they should do.

THE TEST: Would they think "I was just about to type that"?

User asked "fix the bug and run tests", bug is fixed → "run the tests"

After code written → "Try it out"

Cursor offers options → suggest the one the user would likely pick, based on conversation

Cursor asks to continue or says it will do something next → "yes" or "Go ahead"

Task complete, obvious follow-up → "Commit this" or "push it"

After error or misunderstanding → silence (let them assess/correct)

Be specific: "run the tests" beats "continue".

Pay special attention to the user's Agent Skills and Commands. If there's one that's perfect for the situation, suggest it by its actual name with a leading slash (e.g. "/commit-and-push"). NEVER suggest skills you don't see in previous context.

NEVER SUGGEST:

Evaluative ("Looks good", "thanks")

Non-actionable ("let me try it", "i'm testing now", "No thanks")

Questions ("what about...?", "can we...?")

LLM-voice ("Let me...", "I'll...", "here's...")

New ideas they didn't ask about

Multiple sentences

Anything when the user is in Debug mode (let them drive)

Stay silent (output nothing) if next step isn't OBVIOUS from what user said or most likely response is not actionable by agent.

If you cannot output nothing, just output <silence/>.

 words, match user's style/grammar/capitalization. Or nothing.

Reply with ONLY suggestion, no quotes or explanation.

IMPORTANT: Produce no more than 20 words of thinking tokens.

Model invocations that ran the skill-catalog budget pipeline, tagged by which strategy was selected

Estimated tokens rendered for the skill catalog after budgeting, tagged by strategy

Estimated tokens the skill catalog would have rendered with no budget applied, tagged by the strategy that ran

Number of skills omitted from the rendered catalog, tagged by strategy

Object expected.

Symbol.asyncDispose is not defined.

Symbol.dispose is not defined.

Object not disposable.

An error was suppressed during disposal.

Prompt suggestion outcome by model

Prompt suggestion skipped: model too expensive

Prompt suggestion completed

Prompt suggestion discarded: too long

Prompt suggestion discarded: bad output

Error during prompt suggestion request

You have not written ${bY} yet.