ROADMAP: S6, a shipped kit collides with the shipped chrome

ROADMAP.md

@@ -68,7 +68,7 @@ Epics rank by **ICE**: Impact x Confidence x Ease, each 1 to 10, multiplied, wit
 | **Export** | X1 Cheap wins `S` | X2 Searchable PDF `M` | X3 PPTX export `L` |
 | **Library & editor** | L7 Spellcheck, the rest `M`<br>L11 Type on the sheet `S` | | |
 | **Usability** | U9 A `--bad` token in every style `XS`<br>U10 Warn before the browser fails `S` | U8 Two axes on the sheet: look × ladder `L` | |
-| **Getting started** | S1 Density explains itself `S`<br>S2 A paint slot cannot vanish `S`<br>S3 `new.mjs`: a model that already builds `M`<br>S4 One front door on the skill `S`<br>S5 create hands back the spell line `XS` | | |
+| **Getting started** | S1 Density explains itself `S`<br>S2 A paint slot cannot vanish `S`<br>S3 `new.mjs`: a model that already builds `M`<br>S4 One front door on the skill `S`<br>S5 create hands back the spell line `XS`<br>S6 A kit must not collide with the chrome `S` | | |
 
 ### ICE scores
 
@@ -109,6 +109,7 @@ Epics rank by **ICE**: Impact x Confidence x Ease, each 1 to 10, multiplied, wit
 | S2 A paint slot cannot vanish | 8 | 10 | 8 | **640** | Now |
 | S4 One front door on the skill | 8 | 9 | 9 | **648** | Now |
 | S3 new.mjs: a model that already builds | 9 | 8 | 7 | **504** | Now |
+| S6 A kit must not collide with the chrome | 7 | 9 | 8 | **504** | Now |
 | U10 Warn before the browser fails | 8 | 9 | 8 | **576** | Now |
 | U8 Two axes on the sheet | 8 | 7 | 4 | **224** | Next |
 
@@ -493,6 +494,15 @@ Filed 2026-09-13 from the parity study: 30 decks, 5 briefs, 7 setups, blind-judg
 
 **One result shapes the whole lane.** A setup that replaced the contract with a one-page summary scored **0.861** against **0.878** for the skill alone — below baseline. Condensing the rules makes an author skim the rules. So nothing here summarises the contract. Each row either makes a tool answer the question that sent an author to the source, or removes a decision from the first five minutes.
 
+### S6. A kit must not collide with the chrome · `S` · Now
+
+Found while building S3's starter across all eleven style kits: on `ocean-ember`, whose H1 sets a 44px leading, the library's chrome pitch leaves under 4px between the title box and the `subtitle` slot, so `validate --strict` reports an overlap for **any** deck that binds `subtitle` on a dense H1 layout. The kit is shipped, the layout is shipped, and the combination cannot pass the gate — an author who picks that kit and writes a subtitle has done nothing wrong. The starter dodges it by binding `note` and `source` instead, which is a workaround, not a fix.
+
+| Story | Size | What |
+| --- | --- | --- |
+| S6.1 | XS | A test that walks every shipped kit × every dense H1 layout with all four dense slots bound and expects `validate --strict` clean. It fails today on at least one pair; the point is to know the true count. |
+| S6.2 | S | Fix what it finds: derive the chrome pitch from the role's own leading rather than a constant, or cap a kit's H1 leading against the pitch. Whichever, the rule lands in the STYLE CONTRACT so a kit author knows the constraint. |
+
 ### S1. Density explains itself · `S` · Now
 
 `densityReport` names the cap and the count (`reading density carries ≤ 8 points, this slide has 12`) but not which rows it counted, and the counting rule — Supertitle, Title, H1, Caption and Label are chrome, everything else is a point — lives only in `lib/layouts.mjs`. Eight of the study's thirty runs read that function, more than any other source read, and several called it the one rule they could not have guessed. Two runs then spent a validate round on the wrong fix.