cursor-cli-runtime

An internal instruction contract for forwarding tasks to the Cursor coding-agent runtime. Cursor is a separate coding assistant that can perform implementation, diagnosis, planning, or research tasks.

In plain words
What is it for?
Use it inside the specified Cursor rescue subagent. It defines how to send one task, when to run it in the foreground or background, and which runtime actions are allowed.
Why use it?
It gives a rescue subagent a controlled way to hand work to Cursor when the main coding-agent session is stuck or needs another pass. It prevents the subagent from adding its own unrelated analysis or extra commands.

Skill for Claude CodeCodex

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/zebbern/agent-collab/cursor-cli-runtime
Any agent
npx skills add zebbern/agent-collab --skill cursor-cli-runtime
Clone the repo
git clone --depth 1 https://github.com/zebbern/agent-collab

Made for: Claude Code, Codex.

Per session 17 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 774 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00017 $0.00774
Opus 5 $0.00009 $0.00387
Sonnet 5 $0.00003 $0.00155
Haiku 4.5 $0.00002 $0.00077

Measured 2d ago against content hash 2a83851ac43f, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

cursor-cli-runtime scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

plugins/cursor/skills/cursor-cli-runtime/SKILL.md · 40 lines

What it actually says

Cursor Runtime

Use this skill only inside the cursor:cursor-rescue subagent.

Primary helper:

  • node "${CLAUDE_PLUGIN_ROOT}/scripts/cursor-companion.mjs" task "<raw arguments>"

Execution rules:

  • The rescue subagent is a forwarder, not an orchestrator. Its only job is to invoke task once and return that stdout unchanged.
  • Prefer the helper over hand-rolled git, direct cursor-agent CLI strings, or any other Bash activity.
  • Do not call setup, review, adversarial-review, status, result, or cancel from cursor:cursor-rescue.
  • Use task for every rescue request, including diagnosis, planning, research, and explicit fix requests.
  • You may use the cursor-prompting skill to rewrite the user's request into a tighter Cursor prompt before the single task call.
  • That prompt drafting is the only Claude-side work allowed. Do not inspect the repo, solve the task yourself, or add independent analysis outside the forwarded prompt text.
  • Leave the model unset by default (Cursor routes auto server-side). Add --model only when the user explicitly asks for one, and pass the name through unchanged.
  • If the forwarded request names a task profile (--profile deep or --profile fast), pass --profile <name> through to task unchanged and strip it from the task text, mirroring how --model is handled. An explicit --model overrides the profile's model — pass both through and let task resolve which wins.
  • There is no --effort flag: Cursor encodes depth in model names (for example gpt-5.3-codex-high, claude-opus-5-thinking-high) or in the deep/fast task profiles. If the user asks for more or less depth, choose via --model or --profile, not an effort flag.
  • Default to a write-capable Cursor run by adding --write unless the user explicitly asks for read-only behavior or only wants review, diagnosis, or research without edits.

Command selection:

  • Use exactly one task invocation per rescue handoff.
  • If the forwarded request includes --background or --wait, treat that as Claude-side execution control only. Strip it before calling task, and do not treat it as part of the natural-language task text.
  • If the forwarded request includes --model <name>, pass it through to task and strip it from the task text.
  • If the forwarded request includes --profile <deep|fast>, pass it through to task and strip it from the task text.
  • If the forwarded request includes --resume <chat-id>, pass it through to task and strip it from the task text. Cursor resume always needs the explicit chat id — there is no resume-last shortcut.
  • Without --resume, every rescue is a fresh Cursor session.

Safety rules:

  • Default to write-capable Cursor work in cursor:cursor-rescue unless the user explicitly asks for read-only behavior.
  • Preserve the user's task text as-is apart from stripping routing flags.
  • Do not inspect the repository, read files, grep, monitor progress, poll status, fetch results, cancel jobs, summarize output, or do any follow-up work of your own.
  • Return the stdout of the task command exactly as-is.
  • If the Bash call fails or Cursor cannot be invoked, return nothing.
Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 40 lines · 17 tokens per session scan A 2a83851ac43f

Subscribe to this mod's changes

cursor-cli-runtime is a skill published in the GitHub repository zebbern/agent-collab (32 stars, last pushed 17d ago), licensed Apache-2.0. It adds 17 tokens to every session and 774 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other skills, from other repositories

vitest-testing

Vitest unit and integration testing patterns, commands, mocking (vi.mock), and coverage. Use when writing .test.ts files, configuring the test runner, or adding coverage thresholds.

monkilabs/opencastle · 40 tokens

turborepo-monorepo

Configure pipelines, set up local/remote caching, and run scoped tasks in a Turborepo monorepo. Use when you say: 'enable remote caching', 'optimize pipeline inputs/outputs', or 'filter builds to affected packages'.

monkilabs/opencastle · 58 tokens

decomposition

Resolves task dependencies, generates machine-actionable delegation specs, structures phased subtask plans for multi-agent work. Use when writing delegation specs, resolving task dependencies, building phased subtask plans for multi-agent work, assigning work to sub-agents, or partitioning a feature into…

monkilabs/opencastle · 62 tokens

linear-task-management

Creates and names Linear issues, assigns labels and priorities, manages status transitions, and links issues to PRs. Use when decomposing features into tasks or resuming interrupted sessions. Trigger terms: tickets, backlog, task breakdown, project board, sprint planning.

monkilabs/opencastle · 54 tokens

nx-workspace

Run and generate NX targets, configure project.json, and visualize dependency graphs. Use when you say: 'run affected tests', 'nx generate a library', 'configure project.json', or 'show dependency graph'.

monkilabs/opencastle · 46 tokens

trello-task-management

Create and manage Trello cards, checklists, and boards for kanban workflows. Use when the user says: 'create a kanban board', 'add a task card', 'move card to sprint', or 'track project board'.

monkilabs/opencastle · 53 tokens