Skip to content
Skills Library
AI & agents · Tool integration

cursor-cli

/cursor-cli

Use when the orchestrator starts, prompts, reads, or answers a Cursor CLI agent in Herdr — opening Cursor on a project, handing a job agent its task or a skill, reading its output, or handling its workspace-trust dialog.

Licence
MIT
Last repository change
05.11.2026

Assessment and security

100/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content35 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
1
Scripts
0
HTML comments
0
URLs in SKILL.md
0
Deprecation

No signals detected. Repository changed on 05.11.2026, without references to retired models or APIs.

Licence

MIT, open: use and modify while retaining author attribution.

What it does and when to use it

Use when the orchestrator starts, prompts, reads, or answers a Cursor CLI agent in Herdr — opening Cursor on a project, handing a job agent its task or a skill, reading its output, or handling its workspace-trust dialog.

Original author description

Use when the orchestrator starts, prompts, reads, or answers a Cursor CLI agent in Herdr — opening Cursor on a project, handing a job agent its task or a skill, reading its output, or handling its workspace-trust dialog.

How it works

  1. Start
  2. The workspace-trust dialog (read before every first prompt)
  3. Hand it a task
  4. Read state and output
  5. When it's blocked
  6. Stop rules for every job prompt
  7. Quirks and risk
Original SKILL.md headings · 66 lines, 752 words
SKILL.md preview
---
name: cursor-cli
description: Use when the orchestrator starts, prompts, reads, or answers a Cursor CLI agent in Herdr — opening Cursor on a project, handing a job agent its task or a skill, reading its output, or handling its workspace-trust dialog.
---

# Cursor CLI agents

How the orchestrator drives a Cursor CLI (`cursor-agent`) agent running in a Herdr pane. Herdr kind: `cursor`. The `herdr` skill covers Herdr itself; this skill covers what is specific to Cursor.

## Start

- **Interactive (the user takes over):** `bun orch.ts open <project> cursor`. Cursor starts with its normal defaults.
- **Job agent:** `bun orch.ts job spawn <job-id> <project> cursor [--repo-worktree]`. Job agents launch with `--force` (run everything not explicitly denied).
  - In a **job-created worktree** they also get `--trust`. Job worktrees are new directories, and without it every one of them stops at the trust dialog.
  - In **`--repo-worktree` mode** they start in the user's own checkout **without** `--trust`, because that flag would permanently trust the user's checkout. Check that pane for the trust dialog, as for an interactive agent (below).
- If the user's `~/.cursor/cli-config.json` has `approvalMode: "unrestricted"`, the status bar shows **Run Everything** even without `--force`.

## The workspace-trust dialog (read before every first prompt)

In a directory Cursor hasn't trusted, it shows a box: **"⚠ Workspace Trust Required … Do you trust the contents of this directory? [a] Trust this workspace [q] Quit"**.

**Herdr reports this dialog as `idle`, not `blocked`.** Anything you send with `agent prompt` gets typed into the dialog: any `a` in the text trusts the workspace and any `q` quits. So before the first prompt to an interactive (`orch open`) Cursor agent:

```bash
herdr agent read <name> --source visible --lines 40
```

If the screen shows "Workspace Trust Required", don't prompt. Ask the user with AskUserQuestion ("Trust this folder" / "Quit", naming the pane and directory) and send only what they pick: `herdr pane send-keys <pane> a` or `q`. `/open` does this for its own agents. Job agents in job-created worktrees start with `--trust` and never show it; `--repo-worktree` job agents can.

## Hand it a task

```bash
herdr agent prompt <name> "<task text>" --wait --timeout 600000
```

- `--wait` returns at the first settled state. A timeout or `agent_prompt_stalled` doesn't prove the prompt was lost. Read the screen before resending.
- **Pointing it at a skill.** Cursor doesn't load Claude skills, so **always** use the path form. Resolve it with `bun orch.ts skills <project> --resolve <skill>`, then say "Read and follow the skill at `<path>`" followed by the task.
- **Job prompts always end with the stop rules** (below).

## Read state and output

```bash
herdr agent get <name>
herdr agent read <name> --source recent-unwrapped --lines 200
```

A finished reply appears above the `→ Add a follow-up` input line. If a long answer doesn't fit, ask Cursor to write it as Markdown to a temp file and reply with only the path. Use this as a fallback only.

## When it's blocked

Read the screen and tell which it is: a question for the user (route it as the `orchestrate-job` skill says), or an approval prompt. Approval prompts are rare under Run Everything, but a deny rule or the sandbox can still raise one. **Never approve one without the user's explicit instruction.** Show them the exact request and send exactly their answer with `herdr agent send-keys <name> <key>`.

## Stop rules for every job prompt

Paste this block verbatim at the end of every task you give a job agent:

```
Stop rules (from the orchestrator):
- Do not git push and do not open merge requests. Stop and say you are ready to push.
- Do not add or change environment variables, feature flags, secrets, runtime settings, operation contracts, permissions, gates, hooks, or anything that writes to live infrastructure. If you need one, stop and ask.
- Follow this repo's own rules (AGENTS.md, .cursorrules, .cursor/rules, CLAUDE.md) wherever they say more than these.
- When you finish a stage, reply with a short report: what you did, the path of any document you wrote, and any open questions.
```

## Quirks and risk

- It loads `AGENTS.md`, `.cursorrules` and `.cursor/rules` natively, but not `.claude/skills`, so pass skills by path.
- **Nothing hard-blocks `git push`.** With Run Everything / `--force` and an empty `deny` list, Cursor will run any command, so the stop rules are the only guard. If the user wants a hard block, they can add a deny rule for `git push` to `~/.cursor/cli-config.json` themselves. Don't edit it for them.
- The status bar shows the model and the context-use percentage (e.g. `Grok 4.6 Medium Fast · 5%`). A high percentage on a long job is a reason to start a fresh agent for the next stage.

Versions

RepositoryScoreLicence
This page · azis14/herdr-orchestrator100MIT
alexandrbasis/claudops100MIT
alexandrbasis/claudops100MIT

Database snapshot: 13.11.2026 · package version de8967e9b60b (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 100omp-cli
    Use when the orchestrator starts, prompts, reads, or answers an OMP (Oh My Pi) agent in Herdr — opening OMP on a project, handing a job agent its task or a skill, reading its output, or handling it when it is blocked.
  • 90herdr-guide
    Use when the user asks to use Herdr to inspect or control agents, panes, tabs, workspaces, and terminal commands. Triggers on launching agents in Herdr, submitting prompts or native goals, reading agent status or output, and moving running agents between spaces, even when the…
  • 100cursor-custom-prompts
    Create effective custom prompts for Cursor AI. Triggers on "cursor prompts", "prompt engineering cursor", "better cursor prompts", "cursor instructions". Use when working with cursor custom prompts functionality. Trigger with phrases like "cursor custom prompts", "cursor prompts", "cursor".
  • 100cursor-delegate
    Delegate coding tasks to the Cursor Agent CLI (`cursor-agent`) only when the user explicitly requests it, while the orchestrator retains review and landing responsibility.
  • 100cursor-delegate
    Delegate coding tasks to the Cursor Agent CLI (`cursor-agent`) only when the user explicitly requests it, while the orchestrator retains review and landing responsibility.
  • 100herdr-orchestrator
    Coordinate multiple CLI agents in herdr with scoped briefs, verified startup, permission handling, independent review, and monitored handoffs. Use for parallel agent work or observer/evaluator supervision in herdr.

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.