| name | orchestrated-execution |
| description | Use when work needs decomposition, multi-step delivery, or delegation across agents: define execution graph, assign specialized subtasks, and run autonomous fix-and-verify loops until done. |
Skill: Orchestrated Execution
Use this skill to coordinate complex work with deterministic outcomes.
When to Use
- User asks for orchestrator/coordinator behavior.
- Task spans multiple files, systems, or quality gates.
- Work benefits from specialist agents with isolated context.
- Completion requires iterative verification and follow-up fixes.
Core Workflow
- Scope and constraints
- Capture goals, non-goals, and constraints from the request.
- Execution graph
- Split work into ordered phases with dependencies.
- Mark independent tracks for delegation.
- Delegation
- Assign each track to the best-fit specialist agent or skill.
- Integration
- Merge changes and resolve cross-track conflicts.
- Verification loop
- Run deterministic checks.
- If failures exist, create and execute follow-up fixes.
- Repeat until checks pass.
Deterministic Verification Set
Run in this order:
python scripts/verify_agent_harness.py --strict
python validate.py --fast
python -m pytest -q
If a command fails:
- Fix root cause.
- Re-run only affected checks first.
- Re-run the full set before completion.
Delegation Matrix
| Need | Delegate to |
|---|
| External API/docs freshness | Docs Retrieval Expert |
| Setup, missing tools, broken environment | Development Environment Bootstrap |
| Final gate and regression safety pass | Repo Quality Guardian |
| Post-stability simplification/debt reduction | Universal Janitor |
Non-Interactive Delivery Rules
- Continue implementation without asking for confirmation unless a hard blocker exists.
- Prefer reversible, behavior-safe defaults.
- Surface explicit blocker reports only when no safe forward path is available.