task-orchestratorlisted

# task-orchestrator **This is a meta-skill.** It controls HOW other tasks are executed, not WHAT they do. It must be loaded before any non-trivial work begins. ## Pipeline (mandatory for every atomic task) ```mermaid graph LR R["🔍 RESEARCH<br/>audit, gather<br/>evidence"] --> D["🎨 DESIGN<br/>design/optimize<br/>select template"] D --> P["📋 PLAN<br/>numbered steps<br/>verification per step"] P --> E["⚡ EXECUTE<br/>follow plan<br/>record results"] E --> V["✅ VERIFY<br/>check evidence<br/>run CI, tests"] V -->|fail| R V -->|pass| DONE["Done"] ``` ### Phase 1: RESEARCH - Audit current state: read relevant files, check CI logs, inspect git status - Identify all stakeholders: which files, systems, or processes are affected - Gather evidence: do NOT assume, do NOT guess, do NOT rely on memory - Record findings in `.task-orchestrator/research-${task}.md` ### Phase 2: DESIGN - Design the solution approach (or select from existing template) - For known task types: load the subskill template from `skills/.task-orchestrator-templates/` - For novel tasks: design from first principles - If the model/agent suggests a BETTER approach than the template: flag it for template upgrade - Record design decisions in `.task-orchestrator/design-${task}.md` ### Phase 3: PLAN - Decompose into numbered, verifiable steps - Each step must have: action, expected outcome, verification method - If using a template: adapt the template steps to the specific context - Use `updat