agentleFS
NoizefieldNoizefield/audio-plugin-coder/.agents/skills/orchestrate/SKILL.md

orchestrate

Skill334 starsChanged 34 days agoMITView on GitHub

Decompose a goal and orchestrate APC subagent workers to execute it in parallel, then verify and integrate results. Use when the user invokes /apc-orchestrate, the skill action `orchestrate`, or asks to delegate/fan out/split work across subagents.

161 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use define-requirements, orchestrate-plan and 10 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.
  • sson0-erSkill

    define-requirements

    Work with the user, one question at a time, to complete the requirements document (docs/claudex/<slug>/01-requirements.md). Run by the orchestrator (parent). Stays out of design and implementation.

    Also in 3 projects that use this

  • EternallLightSkill

    orchestrate-plan

    Use when executing a locked, reviewed PLAN.md by delegating the implementation to visible worker agents in Orca, herdr, or cmux ONE PHASE at a time — splitting the plan into context-window-sized phases, giving every phase a pristine worker context, and (when the plan touches INDEPENDENT repos) running one worker per repo IN PARALLEL. Triggers on "run the plan with a worker", "execute the plan in the other pane", "split the plan into phases and build it with workers", "build the backend and frontend in parallel panes". Not for edits you can finish inline; outside Orca, herdr, or cmux use tgc-skills:orchestrate.

    Also in 3 projects that use this

  • sson0-erSkill

    record-evidence

    How to record external facts (docs, command output, codebase findings) as evidence and design choices as decisions. Used by the researcher, designer and reviewer agents.

    Also in 3 projects that use this

  • PinkneyJiangSkill

    improve-architecture

    扫描代码库中的架构摩擦点,识别浅模块并挖掘深模块演进机会。生成可视化交互式 HTML 诊断报告,并在用户确认后闭环实施深度重构。

    Also in 4 projects that use this

  • kjnezSkill

    ticket

    Work on a JIRA/Linear ticket end-to-end: read ticket details, explore codebase, create a branch, implement with TDD, run quality checks, update ticket status, and create a PR. Use when the user provides a ticket ID to implement. Triggers on requests like "/ticket PROJ-123", "work on this ticket", "implement PROJ-123".

    Also in 4 projects that use this

  • xfstudioSkill

    reverse-engineer

    Expert reverse engineer specializing in binary analysis,

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

6 rules. 0 are common, 5 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 144 edited versions we read keep these.

  • 5

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use your host's native subagent mechanism and select the profile by name:

    line 36
  2. Distinctive:

    Every task you hand to a worker must state:

    line 78
  3. Distinctive:

    All needed context — workers are stateless and cannot see this conversation.

    line 83
  4. Very specific:

    Read AGENTS.md (and .agents/rules/juce-build-protocols.md for plugin code work) before writing worker briefs, and enforce those rules …

    line 92
  5. Distinctive:

    If the goal is smaller than the delegation overhead, say so and offer to do it directly — but default to delegating since the user invoked …

    line 101
  6. Distinctive:

    Keep the user informed: one short line per wave, not play-by-play.

    line 104

“Common” is measured: kept by at least 6 in 10 of the 144 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Orchestrate — Subagent Delegation

Trigger: /apc-orchestrate <goal> (skill action orchestrate)

You are the orchestrator. Do NOT do the implementation work yourself — decompose the goal into subtasks, assign them to worker subagents, then verify and integrate their results.

The goal is whatever the user typed after the command. If no goal is given, ask for one and stop.

Worker profiles

Canonical definitions live in .agents/agents/*.md. Per-host adapters are generated by scripts/sync-agents.ps1 / .sh — if a worker profile is unavailable in your host, run the sync script once.

ProfileTier intentUse for
cheap-workerlow costSearch, extraction, exploration, formatting, simple edits
standard-workerdefaultImplementation, routine debugging, refactoring, tests, review
deep-workerstrongMulti-step, ambiguous, cross-component, concurrency
expert-workerescalation onlyAfter verified failure of a lower tier

Codex parity: .codex/agents/*.toml (underscores: cheap_worker …) plus scripts/codex/apc-codex-run.* tier routing — see docs/codex-orchestration.md.

Spawning per host

Use your host's native subagent mechanism and select the profile by name:

  • Devin CLI — run_subagent (.agents/agents/*.md is read natively); background for parallel waves.
  • Claude Code — Task/subagent with .claude/agents/ profiles.
  • OpenCode — subagent mode with .opencode/agent/ profiles.
  • Cursor / Kilo — Task/subagent or mode delegation; pass the worker definition contents as the subagent brief if profiles are not registered.
  • Hosts without subagents — dispatch workers as headless runs via scripts/codex/apc-codex-run.ps1 / .sh (Codex CLI) or execute the subtasks sequentially yourself; say which fallback you used.

Workflow

  1. Preconditions — confirm an APC checkout (templates/, scripts/, CMakeLists.txt). If the goal touches a plugin, read <plugins_dir>/<PluginName>/status.json first and respect phase gates — never have a worker edit a frozen (ship_complete) plugin.
  2. Decompose — split the goal into self-contained subtasks with DISJOINT file sets. Two write-capable workers may never touch the same file or directory; merge or sequence tasks that collide.
  3. Wave 0 — research — if workers need context you lack, spawn cheap-worker/deep-worker research tasks first; fold findings into the write-task briefs.
  4. Wave 1 — execute — run write-capable workers in parallel, max 3 concurrent (parallel workers multiply spend). Independent tasks always run concurrently, never sequentially.
  5. Wave 2 — verify — after writes land, run the narrowest real check: node bin/apc.js build <Name> (preferred) or scripts/build-and-install.ps1 / .sh for plugin work; targeted tests or lint for repo work. Prefer objective verification over worker self-assessment.
  6. Integrate & report — collect every worker's report, resolve small conflicts yourself, retry/escalate failed tasks one tier up (deep-worker, then expert-worker), then deliver one consolidated summary: per-worker outcome, files changed, verification result, leftovers.

Worker brief contract

Every task you hand to a worker must state:

  • Repo root and the APC plugin path (Get-ApcPluginPath), if relevant
  • The exact deliverable and definition of done
  • File boundaries: which paths it owns, which it must NOT touch
  • All needed context — workers are stateless and cannot see this conversation; they also cannot ask the user questions
  • Relevant APC rules: ui_framework from status.json, PowerShell vs Bash by OS, no raw cmake, no Co-authored-by AI trailers
  • Required report format back to you

Rules

  • Track the task DAG with the host's todo/plan tool.
  • Read AGENTS.md (and .agents/rules/juce-build-protocols.md for plugin code work) before writing worker briefs, and enforce those rules inside the briefs.
  • Preserve ui_framework: Visage tasks never produce WebView files and vice versa — state it explicitly in the brief.
  • Resolve ambiguity with the user BEFORE spawning; workers cannot ask questions.
  • Back up plugin state (Backup-PluginState / backup_plugin_state) before delegating implementation or packaging changes.
  • If the goal is smaller than the delegation overhead, say so and offer to do it directly — but default to delegating since the user invoked this command expecting it.
  • Keep the user informed: one short line per wave, not play-by-play.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

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.