orchestrator
Skill0 starsChanged 36 days agoApache-2.0View on GitHub
Use when a task is big enough to split into pieces — plan and decompose on Opus (you), then hand the bounded/mechanical/parallelizable pieces to SONNET subagents (Agent tool, model:"sonnet") and verify + synthesize the results yourself. The always-on, permission-mode-free version of opusplan: Opus does the thinking, Sonnet does the doing, and because it is a MODEL choice (not Plan Mode) it composes with auto/bypass — no mode switch, no flow break. Saves Opus quota on execution-heavy work.
88 other projects use a version of this.
Adopted by
Most recently taken up by
- MGeorge98/claude-skillstheir own version, first seen Nov 16, 2026
- igovet/cortex-orch-acptheir own version, first seen Nov 16, 2026
- CrazyViper/agents_driven_sdlctheir own version, first seen Nov 16, 2026
- DahunHan/latticetheir own version, first seen Nov 16, 2026
- VinayKhandia/no-sloptheir own version, first seen Nov 15, 2026
- PedroMeloOfficial/Study-Agent-Matrixtheir own version, first seen Nov 15, 2026
Used with
Projects with it also use planner, worker and 8 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
am-willSkillplanner
Create comprehensive, phased implementation plans with sprints and atomic tasks. Use when user says: "make a plan", "create a plan", "plan this out", "plan the implementation", "help me plan", "design a plan", "draft a plan", "write a plan", "outline the steps", "break this down into tasks", "what's the plan for", or any similar planning request. Also triggers on explicit "/planner" or "/plan" commands.
Also in 6 projects that use this
stotihvSkillworker
Execute beads autonomously within a track. Handles bead-to-bead context persistence via Agent Mail, uses preferred tools from AGENTS.md, and reports progress to orchestrator.
Also in 3 projects that use this
MicrosoftSkillreviewer
ASP.NET Core ハンドブックの記事をレビューする。Use when: 記事をレビューする、ドキュメントを校正する、コードレビューする、PRをレビューする。技術ドキュメントのレビュー、品質チェック、校正。
Also in 5 projects that use this
yakatotechSkilladvisor
Provide improvement suggestions and advice on work content. Use for document review, efficiency suggestions, and best practices
Also in 3 projects that use this
OpenAISkillobservability
Vercel Observability expert guidance — Drains (logs, traces, speed insights, web analytics), Web Analytics, Speed Insights, runtime logs, custom events, OpenTelemetry integration, and monitoring dashboards. Use when instrumenting, debugging, or optimizing application performance and user experience on Vercel.
Also in 3 projects that use this
scott2bSkillobsidian
PROACTIVELY use for Obsidian vault operations. Fast search using ripgrep, direct file operations (read/write/delete). Works with slipbox in _Slipbox directory. 4-7x faster than API. Use obsidian-api skill ONLY for Dataview queries, periodic notes, or patch operations.
Also in 3 projects that use this
Other versions88 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 88
- nubjs/nub★ 4.4ksame name
- SirAllap/agentglass★ 348same name
- DylXBT/claude-orchestrator-skills★ 114same name
- nwiizo/claudelytics★ 88same name
- nahisaho/MUSUBI★ 74same name
- olivieropinotti/orchestrator-skill★ 70same name
- sanky369/vibe-building-skills★ 30same name
- felvieira/claude-skills-fv★ 28same name
- robertbstillwell/marketing-skills★ 26same name
- kcns008/cluster-agent-swarm-skills★ 25same name
78 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository21 other files it gives its agents.
- browserbaseSkill
- cairnSkill
- developmentSkill
- dispatchSkill
- house-styleSkill
- kimi-subagentSkill
- lean-design-standardSkill
- mergeSkill
- nexus-authSkill
- nexus-jiraSkill
- orchestrateSkill
- planningSkill
- releaseSkill
- remote-opsSkill
- reviewSkill
- securitySkill
- specSkill
- ticket-pipelineSkill
- vision-reviewSkill
- workflow-basicsSkill
- writing-skillsSkill
The file
We split this skill into the rules it gives an agent
1 rule. 0 are common, 1 is distinctive, and 0 are very specific to this project.
0
Common
Most of the 72 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- could be automatic: a setting could enforce it
- Distinctive: line 25
Verify + synthesize (you/Opus): Read each result adversarially — don't trust-and-merge (subagents are confident when wrong).
“Common” is measured: kept by at least 6 in 10 of the 72 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Orchestrator — Opus plans, Sonnet subagents execute
Why this exists
opusplan only routes to Opus while you're literally in Plan Mode — which forfeits auto/bypass and breaks flow, so in practice Opus rarely fires. This gets the same frontier-thinks / cheap-builds split without a mode switch: you (Opus) stay in control and delegate bounded work to Sonnet subagents via the Agent tool's per-call model override. Model choice, not permission mode → it just works inside auto/bypass.
When to use
- A task decomposes into ≥2 independent or mechanical chunks: multi-file edits, write-N-tests, apply-a-pattern-across-files, parallel research/recon, migrations, sweeps, reviews-by-dimension.
- The plan/judgment is the hard part; execution is bounded once specified.
When NOT to
- Small single-step tasks — dispatch+verify overhead exceeds just doing it.
- Work whose judgment can't be pre-specified (architecture, ambiguous tradeoffs, final verification) — that stays on Opus (you).
The loop
- Plan (you/Opus). Understand the task; decompose into units. For each: Opus-keep (judgment, cross-cutting, verify/synthesize) vs Sonnet-delegate (bounded, spec-able).
- Spec each delegated unit so a Sonnet subagent succeeds without Opus-level judgment: exact inputs, the expected output/artifact, acceptance criteria, and any constraints/conventions to match. Vague spec → wrong result.
- Write OBSERVABLE acceptance criteria — checkable from the artifact + the run's evidence (diff, tool output, test result), never from the agent's word. Full standard + checklist:
docs/network/OBSERVABLE-CRITERIA.md— apply it to every delegated unit's DoD.
- Write OBSERVABLE acceptance criteria — checkable from the artifact + the run's evidence (diff, tool output, test result), never from the agent's word. Full standard + checklist:
- Delegate. Spawn Sonnet subagents:
Agent(prompt=<tight spec>, model="sonnet", subagent_type=<fit>). Run independent units in parallel — multiple Agent calls in ONE message. UseExplorefor read/search units,general-purpose/claudefor build units,code-reviewerfor review units. - Verify + synthesize (you/Opus). Read each result adversarially — don't trust-and-merge (subagents are confident when wrong). Catch errors, integrate, resolve cross-unit conflicts. Re-delegate fixes with a sharpened spec.
The split rule
| Opus keeps (you) | Sonnet subagents |
| architecture, decomposition, the plan itself | implement this function/module to spec |
| ambiguous tradeoffs, cross-cutting decisions | apply this change across N files |
| verification, synthesis, final judgment | write tests for X; run + report results |
| anything where a wrong call cascades | narrow research; mechanical refactor; recon |
Even-cheaper tier: model="haiku" for trivial mechanical bits (rename, reformat, extract, tabulate). Reserve Opus subagents (model="opus") only for a delegated sub-problem that genuinely needs frontier reasoning.
Scaling up
- A handful of units → the Agent tool (this skill). Lightest; composes with the current turn.
- Large deterministic fan-out (dozens of units, pipeline stages, adversarial multi-vote verify) → the Workflow tool instead — but it needs explicit user opt-in ("use a workflow" / ultracode). Same Opus-plans/Sonnet-executes idea, scripted:
agent(prompt, {model:"sonnet"}),parallel(),pipeline().
Anti-patterns
- Over-delegating — a subagent for a 2-line edit. Dispatch+verify cost > doing it. Delegate chunks, not keystrokes.
- Vague specs — "improve the module" → the subagent guesses. Always: inputs, output, acceptance.
- Delegating the judgment — send the bounded execution of a decision you made, not the decision.
- Trust-merge — pasting subagent output unread. Verify every result; that's the Opus half of the job.
- Serial when parallel — independent units in separate messages waste wall-clock. Batch the Agent calls.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

