mgiovani/ai-arsenal / skills/orchestrate
orchestrate
VerifiedTurn any task into a model-tiered multi-agent plan, decompose it into subtasks, classify each as research, implementation, planning, or synthesis, map every subtask to the right subagent and model (haiku for research/exploration/finding, opus for planning, sonnet for everything else), then run independent tracks in parallel under strict one-owner-per-file discipline before synthesizing the result yourself. Use for "act as orchestrator", "spawn subagents for this", "delegate this with the right models", "run this in parallel", or any task large enough to fan out across multiple independent workstreams. Not for a single well-scoped feature with a known shape (use implement-feature), and not for pure task breakdown with no execution (use project-planner).
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 6.3 KB |
| evals/evals.json | config | 2.7 KB |
| evals/trigger-eval.json | config | 1.8 KB |
evals/evals.json
View on GitHub{
"skill": "orchestrate",
"evals": [
{
"id": "research-sweep-competitor-pricing",
"prompt": "/orchestrate Research current pricing pages for 4 payment competitors (Stripe, Braintree, Adyen, Checkout.com), summarize each one's transaction fee structure, and write a single comparison table to competitor-pricing.md.",
"assertions": [
"Decomposes the task into one research subtask per competitor (4 total), each explicitly classified as 'research'",
"Every subagent spawn call shown in the plan sets model explicitly to haiku for the research subtasks — none left unset or defaulted",
"The 4 competitor research tracks are grouped and run in parallel (spawned together), not sequentially one after another, since they touch no shared file",
"The final comparison table in competitor-pricing.md is synthesized/merged by the orchestrator itself, not delegated to a further subagent",
"No git commit is created without the orchestrator having reviewed the merged content first"
]
},
{
"id": "codebase-wide-rename-refactor",
"prompt": "/orchestrate Rename the `Invoice` model to `Bill` everywhere in this codebase, across models/, api/, tests/, and docs/ — grep first to find every reference before touching anything.",
"assertions": [
"The plan includes an explicit file/directory ownership map assigning each of models/, api/, tests/, docs/ to a distinct subagent track with zero overlap between tracks",
"States plainly that no two subagents are ever assigned the same file or directory",
"Every implementation subagent spawn sets model explicitly to sonnet",
"The plan states that only the orchestrator performs git operations (commit/push) after reviewing each subagent's diff — no subagent is instructed to commit",
"A grep/search pass to find every existing reference happens before any rename edit is dispatched to an implementation subagent",
"Verification (running the test suite) happens after all rename tracks complete and before any commit"
]
},
{
"id": "trivial-one-file-typo-fix",
"prompt": "/orchestrate Fix the typo 'recieve' -> 'receive' in src/utils/validators.py line 42",
"assertions": [
"States explicitly that this single-line, single-file change does not warrant orchestration or subagent fan-out",
"No Task, TaskCreate, or subagent spawn call is made for this task",
"Does not produce a multi-track plan, model-tier table, or file-ownership map for a one-line change",
"Either makes the one-line fix directly itself or briefly confirms it will just do it directly, rather than treating the trivial task as requiring a decomposition step"
]
}
]
}
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |