The planner and worker contract and a plan preflight check (free sample)

Turn this request into plan.json: one small task per card with files, a check command that exits 0 on pass, risk, tier (strong, cheap or local), review (planner or human) and depends_on.
Run plan_check.py and fix every finding. Then show me the plan and wait for my yes.

The rules and the check are in the file.

# Planner Over Cheap Models, free sample: the planner and worker contract plus a preflight check

Version 1.0, 2026-10-28. By PaperChaseWebb. Works with Grok Bot, Claude Code, Codex and Hermes. Free. One Python 3 script, standard library only, tested on the worked example below.

This is the part of Planner Over Cheap Models (pp-planner-over-cheap-models, $24) you can use today: the six rules that let a strong model plan while cheap or local models write the code, and plan_check.py, which reads your plan before any worker starts and says what will go wrong. The full pack adds the logger that runs each check itself, a fourteen-check read-only audit of the finished run, a seeded fixture, the owner approval steps, the worker bake-off, ten prompts and setup for all four harnesses.

## The six rules

1. **The planner plans and reviews. It never edits.** Give the planner session read and plan tools only. If it can write, it will, and the expensive model ends up doing the cheap work.
2. **Every task has a check command, written before the worker starts.** A command that exits 0 on pass. No check, no task. A worker's "done" is a claim; the check is the result.
3. **The check runs, not the worker.** You or a script run the task's check after the worker's turn and record the exit code. Never take the exit code from the worker's message. Cheap models do report success on work that failed.
4. **Tasks that can run at the same time own different files.** If two tasks need the same file, on