Swarm
A portable studio team of reusable AI roles and workflows for Claude, Codex, Gemini, and Grok.
npx -y skills add arclabshq/arc-labs-studio-team --skill swarmAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 20 days oldThe repository was created 20 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Divides a broad problem into bounded expert tracks and merges their evidence into one decision or deliverable. Use when independent research, review, design, or implementation tracks can run in parallel or as sequential specialist lenses.
SKILL.md
1.7 KB, 283 tokens by cl100k_base, as published. Nobody here has run it
Swarm
Use multiple tracks only when decomposition will improve speed or quality. One coherent task with heavy shared context is usually better handled directly.
Orchestration sequence
- Define the shared outcome, constraints, source of truth, and final owner.
- Split the work into non-overlapping tracks with a specific question, input, output, and definition of done.
- Identify dependencies. Run only independent tracks concurrently.
- When the host supports multiple agents and the user or host policy permits delegation, assign bounded tasks without giving every track the whole job.
- When multiple agents are unavailable, run the same tracks sequentially as explicit expert lenses.
- Require each track to return evidence, uncertainty, and recommended action.
- Reconcile disagreements against the shared goal and source of truth.
- Deliver one merged result, not a pile of unreviewed sub-results.
Track contract
Each track should include: question, scope, inputs, exclusions, output format, and completion test. Preserve one owner for synthesis and final quality.
Boundaries
Do not create recursive delegation, duplicate work, or background tasks without a reason. Do not let a track publish, message, spend, delete, or make another external change unless that action is separately authorized. The orchestrator must review all returned work before presenting it as complete.
What ships with it: 1 file
262 B alongside SKILL.md
agents/
- openai.yaml262 B