How to use (60 seconds):
1. Paste the prompt you’re about to send your AI assistant below (optional — it just keeps it in view).
2. Tick every statement that is actually true of your prompt as written — not what you meant; what it says. If a check genuinely does not apply to your task (no money, no shared state), deciding that is a specification act: count it closed — honestly.
3. Read your band and the open failure modes. Close them in the prompt, re-tick, watch Δ shrink.
Note: this page does not auto-analyze your text — you do the scoring, deliberately: judging a prompt against its domain is exactly the skill the book teaches, and no keyword matcher can do it for you. (An AI-powered analyzer is planned; it will be built to the book’s own Chapter 15 security patterns.) Your prompt never leaves your machine.
Your prompt (optional, for reference while you score — it stays on your machine — this page makes no network calls with your text.
Methodology: one point per check, equal weights; verdict bands 0–2 / 3–7 / 8–11 / 12, calibrated against sample prompts. This is a self-assessment heuristic from Delta: Closing the Specification Gap by Sandeep Dhuri (Acuity Press, 2026) — it measures specification completeness, not code quality. Scores are not stored, logged, or transmitted.