agent-orchestration

Getting started with Manager — the 48-tool multi-AI orchestrator. Use when: delegating code tasks to Claude Code/Codex/Gemini, running parallel AI executions, creating workflow templates, or routing tasks across AI backends.

Convene a PAI-style council — 3-round debate where specialists challenge each other. USE WHEN multi-perspective discussion, architecture debate, strategy decisions, cross-domain analysis.

Showcase forge-council — demonstrate the agent roster, council flow, and standalone specialists. USE WHEN demo, showcase, show agents, what can forge-council do.

Convene a developer council — 3-round debate for code review, architecture, and debugging. USE WHEN multi-perspective code review, architecture decisions, team-based problem solving, developer council.

Convene a hiring council -- multi-agent review of job postings, role definitions, and hiring strategy. USE WHEN job posting review, role design, hiring decisions, compensation review, recruitment strategy.

Convene a knowledge management council — 3-round debate on vault organization, memory lifecycle, note architecture, and skill design. USE WHEN knowledge triage, memory promotion, vault organization, note lifecycle, idea graduation, archive decisions.

Convene a product council — multi-agent review of requirements, features, and product strategy. USE WHEN requirements review, feature scoping, product decisions, go/no-go, payments review.

Run OpenAI Codex CLI as a subagent for second opinions, code reviews, and questions. Use when you want a different AI model's perspective.

Run Google Gemini CLI as a subagent for second opinions, code reviews, and questions. Use when you want a different AI model's perspective.

Troubleshoot Claude Code extensions and behavior. Triggers on: debug, troubleshoot, not working, skill not loading, hook not running, agent not found.

Generate PhD-level expert agent prompts for Claude Code. Creates comprehensive 500-1000 line agents with detailed patterns, code examples, and best practices. Triggers on: spawn agent, create agent, generate expert, new agent, agent genesis.

Recommend the right agents and skills for any task. Covers both heavyweight agents (Task tool) and lightweight skills (Skill tool). Triggers on: which agent, which skill, what tool should I use, help me choose, recommend agent, find the right tool.

Detect missing capabilities and route to a subagent with proper access. Use for: missing run_in_terminal, missing shell access, need to execute a PS/bash command, reading process output, script validation, need for a tool unavailable in the current mode.

Calibration and review-pipeline tooling guidance. Use when running or maintaining calibration scripts, review aggregation, and related deterministic assets for issue #360. DO NOT USE FOR: changing orchestration ownership or agent review policy.

Frame credit row emission and deferred credit-input methodology for agents and CE Gate surfaces. Use when an agent or skill needs to write frame credits to the PR-body pipeline-metrics block, or when a pre-PR agent needs to persist a credit-input marker for later harvest. DO NOT…

Guidance-complexity measurement tooling and usage notes. Use when running or maintaining the guidance measurement script and related deterministic analysis assets for issue #360. DO NOT USE FOR: changing D10 ceiling rules or prompt-facing guidance policy, which remain in agent…

Five-scenario workflow troubleshooting guide for diagnosing and fixing common orchestration failure patterns. Use when diagnosing premature implementation, agent confusion, validation gaps, documentation conflicts, or terminal stalls during a workflow session. DO NOT USE FOR:…

Concise reference for deterministic specialist dispatch, review mode routing, CE surface mapping, enum values, and gate criteria extracted from current orchestration agents. Use when an agent or script needs the canonical current routing data without re-deriving it from…

Safe file-operation and issue-creation protocol for Agent Orchestra. Use when choosing workspace tools, avoiding unsafe file writes, or creating GitHub issues under the workflow rules. DO NOT USE FOR: application-level debugging or replacing agent judgment on whether work is in…

Canonical operational contract for Agent Orchestra session-state survival and handoffs. Use when discovering where plan, design, review, startup, calibration, tracking, or plugin state belongs. DO NOT USE FOR: designing new storage mechanisms or writing rationale docs (use…

Guide for creating new skills in this system with proper frontmatter format. Use when adding new skills, updating skill templates, or reviewing skill structure. DO NOT USE FOR: general agent or instruction file customization (use agent-customization) or modifying existing skill…

Canonical agent invocation topology — defines correct sequencing, preconditions, handoff artifacts, and anti-patterns for multi-agent workflows. Use when orchestrating multiple agents or when unsure which agent to invoke next.

Strategies for managing context window saturation in long sessions. Use when the turn counter warns (≥40 turns), when the agent seems to be forgetting earlier instructions, or when starting a continuation session for a multi-day task.

Project-scoped knowledge access protocol — read and write to .github/knowledge/ using standard file tools. Use in every agent that needs to log lessons, decisions, or patterns discovered during a session.

oh-my-opencode를 Copilot CLI용으로 포팅한 멀티에이전트 오케스트레이션 플러그인. Sisyphus(오케스트레이터) + Hephaestus(딥워커) + Prometheus(플래너) + Ralph Loop. "/ultrawork 태스크설명" 으로 시작하면 모든 에이전트가 자동 활성화됩니다.