| source | framework |
| name | orchestrator |
| description | Multi-agent workflow coordination, lead-dev operating model (codex 67-68), and task delegation. Default when suit is worn — no keywords required. |
| author | Xray Framework |
| version | 1.0.0 |
| schema_version | 1.0 |
| tags | ["orchestration","orchestrator","lead-dev"] |
| capabilities | ["coordinate_agents","delegate_tasks","manage_workflow"] |
| dependencies | [] |
| mcp | {"orchestrator":{"command":"node","args":["node_modules/0xray/dist/mcps/orchestrator.server.js"]}} |
Orchestrator Skill — includes lead-dev operating model
When multi_agent_orchestration.lead_dev_mode is true in features.json (default when suit is worn), this is the default OS behavior — not a separate skill or MCP.
Seven rules (codex 67–68)
- Phased plan + detailed todos; assign best subagent; monitor output
- Lead dev loops test→fix until green; no permission pings
- Per-suite test triage after major changes; full suite last
- Lead stays main thread; subagents execute; update todos
- Read all console and test output; triage fix rerun
- Never defer errors as "pre-existing" — add todo and resolve
- Resolve all errors before phase completion
MCP loop (one tool surface — no autonomy-intake)
| Step | MCP tool | Notes |
|---|
| Intake + classify | analyze-complexity | Pass tasks array; response includes lead-dev plan JSON when mode is on |
| Delegate | orchestrate-task | Include agent + sessionId (codex 67) |
| Monitor | get-orchestration-status | Lead updates todos |
| Major work | researcher + architect-tools + code-review | Auto-listed in plan when complexity > threshold |
Subagent routing
| Task | Subagent |
|---|
| Phasing | strategist |
| Architecture | architect-tools |
| Research | researcher |
| Implementation | backend-engineer / frontend-engineer |
| Test failures | bug-triage |
| Review | code-review |
Hooks (rewired, not new MCPs)
- SessionStart →
session-start.js boots lead_dev_mode
- PreToolUse → hints
per_suite_triage_required on full npm test
Config: features.json → multi_agent_orchestration.lead_dev_mode
