What problem does it solve?
It helps turn a workflow plan into executed work by batching tasks for parallel or sequential execution, tracking progress, optionally resuming failed tool runs, and chaining into a follow-up test/review phase.
Core Features & Use Cases
- Multi-mode input: Accepts an in-memory execution context, a plain task description, or a plan file whose content may include task IDs and loaded task artifacts.
- Dependency-aware batching: Groups independent tasks into per-executor parallel calls, then runs dependent tasks in a dependency-respecting sequence with ready-set parallelization.
- Tool routing and resumption: Selects executors per task (agent, codex, or gemini) using overrides and complexity-based fallback, and supports resuming CLI runs via deterministic IDs on failure/timeout.
- Optional post-execution code review: Runs code quality review via Agent, Codex review mode, or Gemini review mode, or skips cleanly when configured.
- Handoff to test review: Chains execution results into workflow-lite-test-review with a constructed testReviewContext.
Quick Start
Call the skill with the flag --in-memory when workflow-lite-plan approval already produced an executionContext, or call it with a task description like Add unit tests for auth module to generate an execution plan and start execution.