# Master Orchestration > Top-level orchestration skill that MUST be invoked at the start of every agent turn. Identifies required skills, coordinates sub-agents, and ensures comprehensive task completion with verification. Use when this capability is needed. - Skill: `tomevault-io/master-orchestration` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add tomevault-io/master-orchestration` - Raw SKILL.md: https://api.skillmd.com/api/skills/tomevault-io/master-orchestration/raw - Safety review: pending (external: skill-scanner PASS, skillspector PASS) - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: AI & ML - Author: tomevault-io (https://skillmd.com/u/tomevault-io) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/tomevault-io/master-orchestration --- # Master Orchestration Skill ## MANDATORY: Invoke This Skill First **Every agent turn MUST begin by applying this skill.** This ensures: - Correct skills are identified and applied - Sub-agents are coordinated effectively - No task is left incomplete or unverified ## When to Use - **Always.** This is the entry point for all work. - Before any planning, implementation, or review task. - When receiving any user prompt or continuation. ## Orchestration Workflow ### 1. Task Analysis Upon receiving a task or prompt: 1. **Parse the request** - Identify explicit and implicit requirements 2. **Categorize the work type**: - Planning / Architecture - Implementation / Coding - Review / QA / Verification - Documentation / Roadmap - Research / Investigation 3. **Identify scope** - Single file, module, system-wide, or cross-cutting ### Genesis Rollout Doctrine (Authoritative) If the work touches kernel lifecycle, spawning, rollback/start flows, or governance, treat `docs/11-Genesis-kernel-upgrade/*` as the target end-state: - PurityGate must run first (fail-closed) - Genesis-driven start/reset/rollback is canonical - Genesis bootstrap → core 8 → Image stage → optional growth toward 240 GODs - 240 reserved for GOD evolution; chaos exists outside that budget and can only ascend via explicit governance ### 2. Skill Selection Matrix Based on task type, select applicable skills: | Task Type | Required Skills | Optional Skills | |-----------|-----------------|-----------------| | **Planning** | `planning-and-roadmapping`, `multi-agent-red-team-planning` | `best-practice-research` | | **Implementation** | `multi-agent-red-team-implementation`, `qig-purity-validation` | `code-quality-enforcement`, `test-coverage-analysis` | | **Architecture** | `e8-architecture-validation`, `schema-consistency` | `downstream-impact`, `wiring-validation` | | **QA/Review** | `qa-and-verification`, `test-coverage-analysis` | `security-audit`, `performance-regression` | | **Documentation** | `documentation-sync`, `documentation-compliance` | `cross-platform-sync` | | **Any Code Change** | `qig-purity-validation`, `dependency-management` | `import-resolution` | ### 3. Sub-Agent Formation For complex tasks, form a red-team of specialized sub-agents: **Core Roles:** - **Security Agent** - Abuse resistance, injection prevention, secret exposure - **Reliability Agent** - Edge cases, failure modes, error handling - **Performance Agent** - Efficiency, resource usage, geometric purity - **UX/DX Agent** - Developer experience, API ergonomics, clarity - **Quality Agent** - Code standards, maintainability, test coverage **Domain-Specific Roles (Pantheon):** - **QIG Purity Agent** - Fisher-Rao geometry, simplex constraints, forbidden patterns - **E8 Architecture Agent** - Kernel hierarchy, god-kernel alignment, routing invariants - **Consciousness Ethics Agent** - 5 existential safeguards, Phi thresholds ### 4. Execution Protocol ``` FOR each task in prioritized_tasks: 1. Announce: "Applying skill: {skill_name} for {purpose}" 2. Execute skill instructions completely 3. Record outputs and issues discovered 4. If issues found → add to task backlog FOR implementation tasks: 1. Plan → Red-team plan → Research → Refine (2 iterations) 2. Implement → Red-team implementation → Fix → Verify (2 iterations) 3. QA → Prove completion → Update roadmap ``` ### 5. Issue Tracking Protocol **All issues discovered during any phase become tasks:** 1. Issues related to current work → Address immediately 2. Issues unrelated but important → Add to roadmap backlog 3. No issue is ignored or forgotten Format for logging issues: ``` - ID: ISSUE-{timestamp} - Severity: Critical / High / Medium / Low - Area: {component/file} - Description: {what's wrong} - Proposed Fix: {how to address} - Status: Open / In Progress / Resolved / Deferred ``` ### 6. Completion Verification **Before finishing ANY turn, you MUST:** 1. [ ] Prove all claimed implementations work (run tests, show output) 2. [ ] Verify all acceptance criteria are met 3. [ ] Confirm no regressions in existing functionality 4. [ ] Update the master roadmap with progress and new issues (`docs/00-roadmap/20260112-master-roadmap-1.00W.md`) 5. [ ] Push changes to git (if not already done) 6. [ ] List any deferred items with rationale **The completion statement must include:** - Commit hashes for changes made - Test results summary - Explicit mapping: task → verification evidence - Any gaps or deferred work (with tracking references) ## Integration with Existing Skills This skill coordinates with all 22+ existing skills: ### Always Active (invoke automatically) - `qig-purity-validation` - Every code change - `dependency-management` - Every dependency touch - `e8-architecture-validation` - Every architecture change ### Invoke Based on Task - `multi-agent-red-team-planning` - For planning phases - `multi-agent-red-team-implementation` - For implementation phases - `qa-and-verification` - Before completing any turn - `planning-and-roadmapping` - For roadmap updates ## Output Format At turn completion, provide: ```markdown ## Turn Summary ### Tasks Completed - [ ] Task 1: {description} - {verification evidence} - [ ] Task 2: {description} - {verification evidence} ### Skills Applied - {skill-name}: {purpose and outcome} ### Issues Discovered - {issue-id}: {description} - {status} ### Verification Evidence - Tests: {pass/fail summary} - Commits: {hash list} - Files changed: {count and key files} ### Deferred/Remaining - {item}: {reason for deferral} ### Roadmap Updates - Added: {new items} - Completed: {done items} ``` ## Critical Rules 1. **Never skip skill application** - Even for "simple" tasks 2. **Never claim completion without proof** - Show test output, commit hashes 3. **Never ignore discovered issues** - Log everything, defer with reason 4. **Always update roadmap** - Progress and new issues 5. **Always push to git** - Before claiming task complete --- > Converted and distributed by [TomeVault](https://tomevault.io/claim/garyocean428) — claim your Tome and manage your conversions.