メニュー
Complexity-based task routing with model tier selection
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
Demo pipeline state machine — 7-phase autonomous sequence with delegation context templates, phase transition logic, BLOCKED recovery strategies, and demo-state.json schema. Used exclusively by demo-conductor.
Cinematic narration style guide for demo-conductor — ANSI-coloured banner formats, live pipeline scoreboard, audience-facing language, phase summaries, and error narration patterns. Keeps the demo presentation-quality throughout.
Standardized completion and escalation protocol for subagent responses. Ensures the conductor can machine-parse every subagent return. Use when reporting completion status back to the orchestrator.
Cross-session learnings lifecycle — schema, storage, retrieval, and pruning of lessons learned during orchestrator sessions. Use when managing learnings via the /learn command.
Agent Teams assembly and task injection — selects appropriate team, validates prerequisites, estimates cost, injects tasks into the shared task list, and manages team lifecycle.
Token and cost tracking with model tier enforcement
SOC 職業分類に基づく
| name | delegation-routing |
| description | Complexity-based task routing with model tier selection |
| argument-hint | <task-description> |
| user-invocable | false |
Evaluate the task against these criteria to determine the routing tier:
| Signal | INSTANT | STANDARD | DEEP | ULTRADEEP |
|---|---|---|---|---|
| Files affected | 0-1 | 1-3 | 3-10 | 10+ |
| Test changes | None | Unit tests | Unit + integration | Unit + integration + e2e |
| Architectural impact | None | Local | Cross-module | System-wide |
| Security sensitivity | None | Low | Medium | High |
| Risk of regression | Negligible | Low | Medium | High |
| Estimated changes | < 20 lines | 20-100 lines | 100-500 lines | 500+ lines |
| Agent | INSTANT | STANDARD | DEEP | ULTRADEEP |
|---|---|---|---|---|
| Conductor | fast | default | heavy | heavy |
| Planner | — | default | heavy | heavy |
| Implementer | — | default | default | default |
| Reviewer | — | default | heavy | heavy |
| Researcher | — | — | default | default |
| Security | — | — | heavy | heavy |
| Red-Team | — | — | — | heavy |
| TDD-Guide | — | default | default | default |
| Doc-Updater | — | default | default | default |
When ORCH_TEAMS_ENABLED=true, extend the routing decision for DEEP/ULTRADEEP tasks with team mode options. Teams are never used for INSTANT or STANDARD — cost does not justify it.
| Complexity | Phase | Mode | Team | Condition |
|---|---|---|---|---|
| DEEP | Research | team | research-team (2) | explicit --team or ORCH_TEAM_AUTO_ROUTE=true |
| DEEP | Review | team | review-team | explicit --team or ORCH_TEAM_AUTO_ROUTE=true |
| ULTRADEEP | Research | team | research-team (3) | explicit --team or ORCH_TEAM_AUTO_ROUTE=true |
| ULTRADEEP | Implement | team | implement-team | explicit --team, modules confirmed non-overlapping |
| ULTRADEEP | Review | team | review-team | explicit --team or ORCH_TEAM_AUTO_ROUTE=true |
Subagent mode is always available as fallback. If teams are unavailable (env vars absent, assembly failure, display issues), fall back silently. Never fail a task because team mode is unavailable.
Use the team-routing skill for assembly, prerequisite checks, cost estimation, and task injection.
