Orca Fleet

Community

larcane97

orca-fleet

Run several coding agents in parallel in Orca worktrees, each under its own clausona profile (another Claude Code or Codex account, or an API model such as GLM or DeepSeek), then watch them, check their work and retire them. Use when the user works in Orca and wants work spread across accounts or models, when one account's limits would slow a fan-out, or says "split this across my accounts in Orca", "run these on DeepSeek in Orca".

Overview

Publisher	larcane97
Repository	`clausona`
Skill name	`orca-fleet`
Stars	57
Forks	11
Bundled files	Instructions only
License	MIT
Links	Skill source Repository Homepage

Markdown instructions
A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.
Works with any LLM
AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.
Self-contained
Everything the model needs lives in the instructions — no extra files to sync.
Open source
Published by larcane97 on GitHub. Read the source before you install it.

Installation

Install the Orca Fleet AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

Open the app and go to Plugins → Skills.
Choose "Install from GitHub".
Paste the skill folder URL below and confirm.
https://github.com/larcane97/clausona/tree/main/plugins/orca-fleet/skills/orca-fleet
Enable the skill in any chat where you want it available.

Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills

git clone --depth 1 https://github.com/larcane97/clausona.git /tmp/clausona
mkdir -p .claude/skills
cp -r /tmp/clausona/plugins/orca-fleet/skills/orca-fleet .claude/skills/orca-fleet

Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Orca Fleet in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Orca Fleet on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Orca Fleet is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

Loaded only when it is needed
The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.
Switch models mid-chat
Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Orca fleet on clausona profiles

Orca runs agents side by side in git worktrees and lists them in its sidebar. clausona puts each terminal's agent on the profile you choose.

0. Load the shared rules

Load the fleet-core skill first and follow it; this skill only adds the Orca steps. In Claude Code, invoke it with the Skill tool. If that fails, or you run in Codex, read the first of these files that exists, where <base> is this skill's base directory:

<base>/../../../fleet-core/skills/fleet-core/SKILL.md
<base>/../../../../fleet-core/*/skills/fleet-core/SKILL.md (if several versions match, read the highest)

If none exists, stop and tell the user to install it: claude plugin install fleet-core@clausona in Claude Code, or codex plugin add fleet-core@clausona in Codex.

1. Before anything

Stop and tell the user what is missing if any of these fails.

orca status shows runtimeReachable: true. If Orca is not open, ask the user to open it.
orca repo list --json lists the repo. If not, ask before adding it with orca repo add --path <repo> --json: it adds the repo to the user's Orca sidebar.
The repo has a remote: workers push their branches.

Pass --json and read handles and paths from the reply; never guess them. Orca runs on macOS, Linux and Windows, and the commands below are the same on each.

Orca has its own account switcher and orchestration commands. Do not use them for workers: clausona run decides each worker's account. Orca also sets CODEX_HOME in its terminals, so a bare codex there runs on Orca's account; clausona run overrides it.

2. Starting a worker

For each task, after fleet-core picked its profile:

Worktree. orca worktree create --repo path:<repo> --name fleet-<task> --json, and record .result.worktree.path (under ~/orca/workspaces/ by default). Then orca worktree set --worktree path:<worktree> --display-name "<task> · <profile>", so the sidebar shows which account runs where. Keep it under 26 characters.
Terminal. orca terminal create --worktree path:<worktree> --json, and record .result.terminal.handle. Do not pass --command: the shell's start-up output can swallow it. Do not rely on --title either: the shell renames the tab.
Wait for the shell. Read orca terminal show --terminal <handle> --json once a second until its lastOutputAt is at least two seconds old, for at most 20 seconds. orca terminal wait does not work for a plain shell.
Start. orca terminal send --terminal <handle> --text "clausona run <profile> -- <args>" --enter, with the arguments from fleet-core section 4. Codex workers keep --disable hooks.
Check it started. After about 10 seconds, run orca terminal show --terminal <handle> --json. The tool has started when .result.terminal.title is no longer the shell's (Claude Code sets ✳ Claude Code), or when orca worktree ps --json lists an agent under the worktree. The preview field shows the screen, but it can lag behind a full-screen app that is still starting, so never resend the launch line on the preview alone: a line sent to a running Claude Code becomes its first prompt. Send it once more only when the title is still the shell's after 20 seconds and the preview shows the launch line never ran. Otherwise tell the user what the screen shows.
Folder trust. The first worker of a profile in a repo that profile never trusted asks whether to trust the folder. Answer only that question, and only about this repository:
- Claude Code ("Is this a project you created or one you trust?"): the highlighted answer is "No, exit". Send the down arrow, orca terminal send --terminal <handle> --text $'\e[B' (PowerShell: --text "$([char]27)[B"), then orca terminal send --terminal <handle> --enter.
- Codex ("Trust this folder?"): the highlighted answer is "1. Trust and continue". Send orca terminal send --terminal <handle> --enter.
Brief. Write it as fleet-core section 5 says, then orca terminal send --terminal <handle> --text "Read .fleet-brief.md in this folder and do what it says." --enter, and start waiting on it (section 3). In Codex, wait only once every worker has its brief.
Record the handle, worktree, branch and profile in the table.

Start workers on the same Claude Code profile one at a time, each after the previous one shows its first output.

The user can open any worker's tab in Orca, read along and type into it at any time.

3. Waiting and reading

Wait: right after each send, run sleep 5; orca terminal wait --terminal <handle> --for tui-idle --timeout-ms 1800000 (PowerShell: Start-Sleep 5; orca terminal wait …). In Claude Code, run it as a background task. The pause matters: right after a send the worker has not started, its screen is still idle, and tui-idle would return at once. The wait returns when the worker's screen goes idle: its turn ended, or it waits at a prompt.
Check it really stopped: for a Claude Code worker, orca worktree ps --json lists the agent under its worktree with state and lastAssistantMessage, from Orca's hooks (clausona shares them across profiles). If state is still working, wait again without the pause. Codex workers do not appear there; read their screen.
The reply: lastAssistantMessage is a Claude Code worker's last reply, up to its first 8000 characters. Look for the DONE or BLOCKED line there first. When the reply is longer, its last line is cut off: then, or when the field is missing, use fleet-core's last-reply.mjs (fleet-core section 8).
Read: orca terminal show --terminal <handle> --json, field .result.terminal.preview. It shows the current screen, including Claude Code's and Codex's full-screen views, once they are up; orca terminal read returns only scrollback.
Follow-up: write the file, then orca terminal send --terminal <handle> --text "Read .fleet-followup-<n>.md in this folder and do what it says." --enter, and wait as above.
After a timeout: read the screen. If the worker still works, wait again without the pause.
In Codex (one foreground wait at a time, three minutes each): give every worker its brief before you wait on any, then wait on each unfinished worker in turn with orca terminal wait --terminal <handle> --for tui-idle --timeout-ms 180000. Keep the five-second pause only for a worker you sent something to in the last five seconds.

Orca has no "until it works again" wait. For a worker the user is answering, follow fleet-core section 7.

4. When an account runs out

Create a new terminal in the same worktree and start the new profile there, as in section 2, steps 2 to 7, with fleet-core section 9's brief.
orca terminal close --terminal <old handle>.
orca worktree set --worktree path:<worktree> --display-name "<task> · <new profile>", so the sidebar names the new account. Keep it under 26 characters, as in section 2.

5. Retiring

After fleet-core section 11's checks, run orca worktree rm --worktree path:<worktree> --json. Never add --force. Orca refuses a worktree with changes ("Failed to delete worktree"); then keep the worker and report why. When it succeeds, Orca closes the worktree's terminals and deletes its local branch too, so only the pushed branch remains: never retire before the push check. Confirm the worktree directory is gone.

Never close terminals or remove worktrees you did not create.

Frequently asked questions

What does the Orca Fleet AI skill do?

Why use Orca Fleet on TypingMind?

Because you install it once and use it with any model. Orca Fleet is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Orca Fleet in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/larcane97/clausona/tree/main/plugins/orca-fleet/skills/orca-fleet. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Orca Fleet?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Orca Fleet?

As many as you like. As long as a model supports skills, you can use Orca Fleet with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Orca Fleet AI skill free?

Yes. It is published on GitHub by larcane97 under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Openclaw Live Updater

openclaw

OrganizationPopular

Maintain the canonical live OpenClaw main checkout, macOS LaunchAgent-managed Gateway, local macOS app, exact-head main CI, and recurring full release validation. Use for fast-forward update…

Openclaw Pr Maintainer

openclaw

OrganizationPopular

Review, triage, repair, or land OpenClaw issues and pull requests with current-source evidence and the native maintainer workflow.

Openclaw Refactor Docs

openclaw

OrganizationPopular

Refactor an existing OpenClaw docs page with source-audited preservation, restructuring, and verification.

Openclaw Repair Sweep

openclaw

OrganizationPopular

Run scoped OpenClaw issue/PR repair campaigns: coordinate workers, prove root causes, and land or close verified work under the requested authority.

Overview

Installation

Install in TypingMind

Install in another agent

Use it in TypingMind

Works with any AI model

Skill instructions

Orca fleet on clausona profiles

0. Load the shared rules

1. Before anything

2. Starting a worker

3. Waiting and reading

4. When an account runs out

5. Retiring

Frequently asked questions

Related AI Skills

Set up your own AI workspace now