What problem does it solve?
It prevents agent workflows from skipping planning by converting a multi-outcome user request into clear control artifacts (spec, plan, and result checks) that downstream skills can execute safely and coherently.
Core Features & Use Cases
- Compound request detection: Ensures the skill only runs when multiple deliverables, pillars, dependencies, or gates are involved.
- Durable workbench artifacts: Creates
spec.md, plan.md, and result-check.md under project/workbench/specs/<slug>/ for stable downstream execution.
- Gate-and-blocker visibility: Keeps upstream evidence and decision requirements explicit, marking dependent deliverables as blocked when prerequisites are missing.
- Separation of concerns: Enforces that raw sources stay in
project/sources/, synthesis stays in project/workbench/, and completed outputs go to project/artifacts/, without writing into project/brain/ unless a type: decision is properly logged.
Quick Start
Ask for something like an SEO technical audit plus a content plan plus publication readiness review, and instruct the agent to create a durable spec and plan for the SEO deliverables while marking website generation as out of scope.