Calibration Debate

# Calibration Debate

> Turns the assistant into a debate panel plus a judge: takes a thesis or a prior call, argues both sides as hard as it can (a real steelman for and against), then scores how defensible the original reasoning was on a 1-5 rubric — a calibration layer that tells you whether a decision was sound, separate from whether it happened to work out.

| | |
|---|---|
| **Use when** | You want to pressure-test a thesis, a recommendation, or a past call — before committing, or as a post-mortem on decision quality |
| **Produces** | A steelmanned pro case, a steelmanned con case, a five-dimension verdict, a 0-100 defensibility score, and the single biggest lesson |
| **Depth** | Medium — a structured debate and verdict |
| **Pairs with** | [`prompts/research/idea-generation.md`](idea-generation.md) · [`prompts/research/cross-source-synthesis.md`](cross-source-synthesis.md) |
| **Run-time needs** | **None — the prompt block below is fully self-contained.** For the strict voice + a Word / Excel / PDF / HTML deliverable, also attach [`BASE.md`](../../BASE.md) — one prompt + `BASE.md`, never a third file. |

---

## The prompt

Copy everything in the block below. Replace the `{{PLACEHOLDERS}}` before sending.