recursive-10-10-product-hardeninglisted

# Recursive 10/10 Product Hardening Drive a product from current reality to **evidence-backed 10/10** by deriving a behavior spec *from the code*, then continuously dogfooding, fixing, and re-verifying the gap between what the code actually does and what it should do. The output is not a cleaner UI or a report — it is a **known-good product state plus a current behavior-spec workbook** that maps every feature to the evidence proving it works. **Why a code-derived behavior spec, not a checklist:** a checklist tests what someone *remembered* to write down; a code-derived spec tests what the product *actually claims to do*, feature by feature, traced back to the source. The gap between "the code has a path for this" and "a real persona can complete it, verified" is where products quietly fail — and that gap is exactly what this skill catalogs, then closes. ## Core rule Do not claim 10/10 from inspection, intent, wording, or partial tests. Claim it only when the current code, the behavior-spec workbook, browser evidence, and automated test results **all agree**. **Success is shown as evidence, never asserted** — never report a row Verified without the supporting output. ## When to use - You want a product driven to genuine, defensible 10/10 — not flagged, *fixed and proven*. - You need a canonical, code-grounded behavior spec: every feature → its expected behavior from the source → tested → fixed → verified. - You want an autonomous test→fix→retest loop that keeps going unt