from orchestrate17
Supervisor skill that composes multi-model runs, routes subtasks to the right models and skills, and manages implementation & verification runs.
Orchestrate is a supervisor skill for composing multi-model workflows. It helps agents split a large task into verified implementation and review runs, pick appropriate models, load and combine skills, and run verification (tests/smoke checks) before concluding. The skill emphasizes step boundaries, verification ownership, and reuse of local runner scripts (run-agent.sh, run-index.sh).
Use Orchestrate when a task requires multiple specialist runs (implementation + reviewers), when you need to fan-out reviews across different model families, or when a change must be implemented, tested, and re-reviewed systematically. Also useful for complex research-to-implementation workflows where verification and staged steps are required.
Designed for agents and runtimes that can execute shell-run scripts and subagent runs (CLI-driven models and local orchestrators). Works well with Copilot/Codex-style tooling and any agent that can call run-agent.sh or equivalent orchestration CLIs.
This skill has not been reviewed by our automated audit pipeline yet.
