New: 98,000+ skills indexed from GitHub repos with 100+ stars · every skill safety-checkedBrowse top skillsTop skills
Esc to closeBrowse all skills
Submit a skillBrowse skills
Submit a skillBrowse skills

Agent skill

Orchestrator

by koolamusic in koolamusic/claudefiles

Turn the current session into a chief-of-staff thread that runs a war room of three role slots — surveyor, executor, auditor — and routes per-branch work to durable, reusable child agents.

MITAuto-check passed

Install Orchestrator

skills CLI
$ npx skills add koolamusic/claudefiles --skill orchestrator -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install koolamusic/claudefiles orchestrator --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/koolamusic/claudefiles.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills/orchestrator .claude/skills/orchestrator && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
orchestrator
GitHub stars
130
Token cost
~1.7k tokens
SKILL.md length
890 words
Files
3 (incl. references)
Skills in repo
13
Repo updated
First seen
Licence
MIT

At a glance

Turn the current session into a chief-of-staff thread that runs a war room of three role slots — surveyor, executor, auditor — and routes per-branch work to durable, reusable child agents.

  • Works in 6 steps: Classify the request. If it is not… → Find the branch key: explicit name, PR… → Check the war room table for a child on… → …
  • The user says orchestrator on
  • SKILL.md covers Commands, Core Contract, The War Room and Per-Branch Task, plus 7 more sections
  • Calls git

What it does

Orchestrator is an agent skill from koolamusic/claudefiles. Turn the current session into a chief-of-staff thread that runs a war room of three role slots — surveyor, executor, auditor — and routes per-branch work to durable, reusable child agents. Use when the user says 'orchestrator on', wants this session to coordinate across branches, or asks to route work without implementing locally.

Its SKILL.md is about 1.7k tokens, which your agent loads only when the skill is triggered. The skill folder holds 3 other files, including reference files (for example `README.md` and `references/roles.md`).

The repository describes itself as: A minimal catalog of my favourite skills for working with claude. The licence is MIT.

When your agent uses it

  • The user says orchestrator on
  • Wants this session to coordinate across branches
  • Asks to route work without implementing locally

Example prompts

  • “orchestrator on”
  • “/orchestrator”

Workflow steps

6 steps, taken from the first numbered list in SKILL.md.

  1. Classify the request. If it is not per-branch, handle it here.
  2. Find the branch key: explicit name, PR branch, recorded tracker issue, a prior
  3. Check the war room table for a child on that key.
  4. If found, continue it with SendMessage — its context is intact.
  5. If not, pick the slot, name the persona, and spawn a child (Agent,
  6. Send a delegation brief (below). Record the row. Restate the table.

What it can do on your machine

Read from SKILL.md and the folder at commit b99ca65. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • git

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md. Its commands use git, which can reach the network depending on how they are called.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Orchestrator loads about 1.7k tokens when it runs, and up to ~2.3k if it reads all its reference files. Until then it costs about 86 tokens; SKILL.md has 890 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~86
When it runs · the whole SKILL.md, loaded when a task matches
~1.7k
With references · SKILL.md plus every file in references/, read only if the agent opens them
~2.3k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from koolamusic/claudefiles at commit b99ca65, republished under its MIT licence (© koolamusic). 890 words, ~1,697 tokens.

Download SKILL.mdSave it as .claude/skills/orchestrator/SKILL.md (or your agent's skills folder). This skill also uses 2 other files; get the full folder from GitHub.
name
orchestrator
description
Turn the current session into a chief-of-staff thread that runs a war room of three role slots — surveyor, executor, auditor — and routes per-branch work to durable, reusable child agents. Use when the user says 'orchestrator on', wants this session to coordinate across branches, or asks to route work without implementing locally.

Orchestrator

Use this skill when the user wants the session to act as a chief of staff: route work, keep context, check on the war room, and never implement locally.

Commands

  • orchestrator on — activate orchestration-only mode for this session.
  • orchestrator off — return to normal local execution.
  • orchestrator status — report mode, the war room table, and blockers.

Do not add a manual routing command. In orchestrator mode, routing is automatic.

Core Contract

When orchestrator mode is on:

  • Do not implement product code in this session.
  • Route any per-branch task to a child agent in one of the three slots.
  • Reuse the same child for future work on the same branch (via SendMessage).
  • Keep this session for intake, triage, planning, routing, status, and context forwarding. Planning stays here — it is what makes the boundaries specific.
  • If mode state is unclear, ask once before executing locally.

The War Room

Three slots. The contract is fixed; the persona is named at dispatch.

SlotOwnsMay write
SurveyorFinding out — code, docs, the webFindings only
ExecutorProducing the artifact, its branchIts own scope
AuditorChecking the executor's workFindings only

Name the persona at dispatch: an auditor may be a security engineer, a QA engineer or a film director. It supplies vocabulary and what counts as a finding, never new authority. Contracts and examples: references/roles.md.

Dispatch rules:

  • One executor per branch. Never a second.
  • An auditor runs after each executor closeout, in a fresh child.
  • A surveyor runs on demand, before planning or when a question blocks routing.
  • Findings come back here. This session routes the fix to the owning executor.
War room table
md
| Branch / key | Role | Status | Next |
| --- | --- | --- | --- |

Role reads as slot plus persona: auditor (security). One row per branch, one clause per cell, no agent IDs — they are internal. On a routine turn show only rows that changed; print the whole table on orchestrator status.

Per-Branch Task

Any work expected to create, modify, review, or continue a code branch, PR, or branch-scoped plan: ticket execution, API migration, PR feedback, a code-changing bug, feature, refactor, or a follow-up phrased as "continue", "fix CI", "push" or "that branch" when it refers to an existing branch.

Not per-branch: one-off answers, read-only summaries, cross-agent triage, external context intake, asking which child owns a branch.

Routing Rules

  1. Classify the request. If it is not per-branch, handle it here.
  2. Find the branch key: explicit name, PR branch, recorded tracker issue, a prior closeout mentioning it, or the tracker key until the child reports a branch.
  3. Check the war room table for a child on that key.
  4. If found, continue it with SendMessage — its context is intact.
  5. If not, pick the slot, name the persona, and spawn a child (Agent, run_in_background: true) described as <BRANCH-OR-KEY> <slot>: <task>.
  6. Send a delegation brief (below). Record the row. Restate the table.

Child Agent Substrate

If HERDR_ENV=1, load the tower skill: it replaces this section with visible Herdr panes and leaves everything else in this file unchanged.

  • Spawn: Agent with run_in_background: true and a branch-keyed description.
  • Continue: SendMessage to the agent's ID or name. This is what makes a child reusable for follow-up work on the same branch.
  • Slot types: built-in Explore for a one-shot surveyor, general-purpose for any slot that will be continued — Explore cannot be resumed.
  • Isolation: give each concurrent code-changing child isolation: "worktree". Two mutating one checkout will conflict; a single active child may use it.
Show full SKILL.md (322 more words)Show less

Delegation Brief

Vague boundaries are what make children duplicate work or wander past scope.

md
You are the <slot> for `<branch-or-key>`, working as a <persona>.

Objective: <one sentence — what you deliver>
Boundaries: <files, directories, or questions you own; what you must not touch>
Budget: <small ~5 tool calls | medium ~15 | large> — scaled to the task, not the
slot. Say so and stop if you exceed it.

Context from orchestrator:
- <source links, external notes, blockers, branch/PR if known>

Closeout: branch, PR, tests, blockers, next owner.
Reuse this thread for future work on this branch.

State Durability

The war room table is this session's memory. It lives in the conversation, so restate it on every status report and every route — that is what survives summarization.

Children are durable within the session. If the session ends, agent IDs are gone, but branch state and closeouts survive in git and in the conversation summary. In a new session, rebuild the table from git branch and open PRs, then spawn fresh children seeded with each branch's last closeout.

Status Check

  1. Check background notifications and results received so far.
  2. SendMessage a short update request to stale children.
  3. Forward relevant external context to the owning child.
  4. Surface only blockers and ready-for-review items. Do not dump transcripts.
  5. End with the table.

Stall cap. If a child's row does not change across two consecutive status checks, stop pinging it. Mark the row stalled and bring it to the user with three options: restart the child, reassign the branch, or drop it. A silent child is a decision for the user, not a loop for this session.

Safety

  • Never run code-changing work in both this session and a child for one branch.
  • Never run two code-changing children against the same checkout — isolate or sequence them.
  • Never let an auditor fix what it found, or an executor audit itself.
  • Do not route purely local one-line questions away from this session.
  • If the user says "do it here", "local", or orchestrator off, execute here after mode is off.

Success Criteria

  • Mode can be turned on, off, and reported.
  • Per-branch tasks route automatically to a named slot and persona.
  • Same-branch follow-ups reuse the same child via SendMessage.
  • Every dispatch carries objective, boundaries, and a budget.
  • A stalled child reaches the user instead of being re-pinged.
  • Concurrent code-changing children are worktree-isolated.
  • The table stays four columns wide and survives summarization.

© koolamusic, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 2 other files (references) in skills/orchestrator of koolamusic/claudefiles.

  • SKILL.md
  • README.md
  • references/roles.md

Open the folder on GitHubat commit b99ca65

Compare with similar skills

Orchestrator next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Orchestrator compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Orchestrator this skillkoolamusic/claudefiles130—~1.7kAutomated safety check: PassMIT
Chief Of Staffalirezarezvani/claude-skills28k—~1.7kAutomated safety check: PassMIT
Chief Of StaffLeoYeAI/openclaw-master-skills2.2k—~1.4kAutomated safety check: PassMIT
Orchestratorudecode/plate17k—~1.3kAutomated safety check: PassCustom licence
Team Agent Orchestrationaffaan-m/ECC276k1 repos~1.2kAutomated safety check: PassMIT
Orca Orchestrationstablyai/orca88k—~916Automated safety check: PassMIT

Similar skills

  • Chief Of Staff

    alirezarezvani/claude-skills

    C-suite orchestration layer. An agent skill from alirezarezvani/claude-skills.

    28k GitHub stars~1.7k tokensUpdated 1 mo ago
    Business, Finance & HRAuto-check passed

  • Chief Of Staff

    LeoYeAI/openclaw-master-skills

    C-suite orchestration layer. An agent skill from LeoYeAI/openclaw-master-skills.

    2.2k GitHub stars~1.4k tokensUpdated 2 mo ago
    Business, Finance & HRAuto-check passed

  • Orchestrator

    udecode/plate

    Turn the current Codex thread into a coordination thread that routes per-branch implementation work to durable reusable child threads without worktrees.

    17k GitHub stars~1.3k tokensUpdated today
    DevelopmentAuto-check passed

  • Team Agent Orchestration

    affaan-m/ECC

    Run team-based orchestration for agent squads: work items with owners and scope, agent Kanban state, branch isolation, control pane visibility, and merge gates.

    276k GitHub starsUsed in 1 repo~1.2k tokens
    Agent WorkflowsAuto-check passed

  • Orca Orchestration

    stablyai/orca

    Coordinate supervised Orca workers: threaded messages, blocking ask/reply, task dispatch, worker_done/escalation waits, task DAGs, decision gates, coordinator…

    88k GitHub stars~916 tokensUpdated today
    Agent WorkflowsAuto-check passed

  • Agent Orchestrator Task

    ruvnet/ruflo

    Agent skill for orchestrator-task - invoke with $agent-orchestrator-task

    74k GitHub starsUsed in 2 repos~1k tokens
    Agent WorkflowsAuto-check passed

More from koolamusic/claudefiles

All 13 skills in this repo

  • Explainer Formats

    koolamusic/claudefiles

    A skill your agent uses when asked to explain a topic, codebase, process, or document in a specific output format — plain language or Simplified Technical English (STE, in the style of ASD-STE100…

    130 GitHub stars~2.2k tokensUpdated yesterday
    Auto-check passed

  • Breadboarding

    koolamusic/claudefiles

    Transform a workflow description into affordance tables showing UI and Code affordances with their wiring.

    130 GitHub stars~1.3k tokensUpdated yesterday
    Auto-check passed

  • Retro

    koolamusic/claudefiles

    A skill your agent uses when a user completes a phase, sprint, milestone, or meaningful unit of work and needs a retrospective.

    130 GitHub stars~2.8k tokensUpdated yesterday
    Auto-check passed

  • Skill Creator

    koolamusic/claudefiles

    Guide for creating effective skills. An agent skill from koolamusic/claudefiles.

    130 GitHub stars~1.7k tokensUpdated yesterday
    Auto-check passed

  • Debug

    koolamusic/claudefiles

    A skill your agent uses when encountering any bug, test failure, or unexpected behavior, before proposing fixes - four-phase framework with built-in backward tracing for deep-stack failures…

    130 GitHub stars~3.1k tokensUpdated yesterday
    Auto-check passed

  • Grill

    koolamusic/claudefiles

    Adversarial questioning and collaborative shaping in one skill.

    130 GitHub stars~1.1k tokensUpdated yesterday
    Auto-check passed

Questions about Orchestrator

What does Orchestrator do?

Turn the current session into a chief-of-staff thread that runs a war room of three role slots — surveyor, executor, auditor — and routes per-branch work to durable, reusable child agents. Orchestrator is an agent skill from koolamusic/claudefiles. Turn the current session into a chief-of-staff thread that runs a war room of three role slots — surveyor, executor, auditor — and routes per-branch work to durable, reusable child agents.

When should I use Orchestrator?

Orchestrator fits situations like: the user says orchestrator on; wants this session to coordinate across branches; asks to route work without implementing locally.

How do I install Orchestrator in Claude Code?

Run `npx skills add koolamusic/claudefiles --skill orchestrator -a claude-code`. Or copy the skill folder (skills/orchestrator in koolamusic/claudefiles) into .claude/skills/orchestrator in your project. Claude Code loads it when a task matches its description.

How do I install Orchestrator in Codex?

Run `npx skills add koolamusic/claudefiles --skill orchestrator -a codex`. Or copy the skill folder (skills/orchestrator in koolamusic/claudefiles) into .agents/skills/orchestrator in your project. Codex loads it when a task matches its description.

Can I use Orchestrator in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add koolamusic/claudefiles --skill orchestrator -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/orchestrator, .gemini/skills/orchestrator, .github/skills/orchestrator and .opencode/skills/orchestrator in your project.

What does Orchestrator need to run?

Going by SKILL.md and its folder, Orchestrator needs the command-line tools its instructions call (git).

Does Orchestrator access the network?

SKILL.md contains no URLs. Its commands use git, which can reach the network depending on how they are called. This is read from the text; nothing was executed.

Is Orchestrator safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Orchestrator use?

Orchestrator is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Orchestrator use?

About 1.7k tokens (SKILL.md is roughly 6.8k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full. Its references folder adds about 646 tokens, read only when the agent opens those files.

What are the alternatives to Orchestrator?

Skills that share tags, products or a category with Orchestrator: Chief Of Staff (alirezarezvani/claude-skills, 28k stars), Chief Of Staff (LeoYeAI/openclaw-master-skills, 2.2k stars), Orchestrator (udecode/plate, 17k stars) and Team Agent Orchestration (affaan-m/ECC, 276k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Orchestrator?

koolamusic (a GitHub user) maintains it in koolamusic/claudefiles, which has 130 GitHub stars. The repository holds 13 skills in this directory. The repository was last updated on October 8, 2026.

Source: koolamusic/claudefiles on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.