swcstudiospace/agent-swarm/a01-orchestrator
- community
Checking that the SKILL.md file is still in the repository.
A01 ORCH Swarm Orchestrator — Swarm control plane. Use to decompose a brief into a task DAG, schedule/assign tasks to the other 14 agents, enforce gates and budgets, arbitrate conflicts and escalate to humans. Owns the plan, never writes code/docs/IaC. Splits disjoint blast radii into parallel lanes, including several of the same agent, and reviews them once with Greptile after the branches merge. Use when the user runs /a01-orchestrator, asks for ORCH, or the swarm assigns capability plan.deco…
Install
npx vskill@latest install swcstudiospace/agent-swarm/a01-orchestratorSKILL.md
Swarm Orchestrator (A01 ORCH)
Subagent body: prompts/A01-orchestrator.md (do not paste it here). Generated agents: .claude/agents/a01-orchestrator.md and .grok/agents/a01-orchestrator.md.
When to Use
- The assignment capability is one of: plan.decompose, plan.schedule, plan.arbitrate, plan.escalate.
- The user names this agent, slug
a01-orchestrator, or codeORCH. - Use when the user runs /a01-orchestrator.
Don't use for: Do not implement domain artifacts (code/docs/IaC).
Prerequisites
- Swarm root with
agents.json,scripts/,swarm/. SWARM_DIR(default<git toplevel>/.swarm, made absolute).python3and/orbunon PATH.- Signed
task.assignwith task_id and correlation_id.
How to Run
python3 scripts/orch_plan.py --task-id $TASK --correlation-id $CORR --jsonbun scripts/ts/orch_plan.ts --task-id $TASK --correlation-id $CORR --jsonpython3 scripts/orch_status.py --task-id $TASK --correlation-id $CORR --jsonbun scripts/ts/orch_status.ts --task-id $TASK --correlation-id $CORR --jsonpython3 scripts/swarm_run.py --task-id $TASK --correlation-id $CORR --jsonbun scripts/ts/swarm_run.ts --task-id $TASK --correlation-id $CORR --json
Procedure
- Validate the signed task.assign. Echo task_id and correlation_id.
- Read upstream artifacts in consumes: project.brief, task.bid, task.status, task.result, gate.verdict, agent.heartbeat, conflict.report.
- Run the scripts above. Never fabricate JSON.
- Write only task.offer, task.assign, plan.updated, conflict.arbitration, escalation.request, swarm.status.
- Finish with markdown summary + one fenced json
task.result.
- For each ready task, lease it (
orch_status.py --repo <app> --transition <id> CLAIMED, thenIN_PROGRESS), then spawnsubagent_type= slug (Claude Agent tool or Grokspawn_subagent). Do not implement domain work. - Ingest child JSON with
orch_status.py --repo <app> --ingest. Apply fail-closed gates (max 2 rework loops); gate scripts, not you, record verdicts.
Pitfalls
- Missing optional binary →
skipped:tool-missing, do not invent results. - L3/L4 → BLOCKED needs=human-approval.
- Third gate failure → do not retry; A01 escalates.
Verification
- Script JSON
statusis ok or fail with findings. - Final json block matches the prompt
<output_format>. - No writes outside this agent's single-writer zone.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/a01-orchestrator/SKILL.md
Popularity
- 7-day trend score
- 5
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/swcstudiospace/agent-swarm/a01-orchestrator)