PR-1: clean-slate v2 rail + qtest false-green fix

CLAUDE.md

@@ -79,7 +79,7 @@ Operational knowledge lives in `.okf/` (`build/`, `content/`, `design/`, `workfl
 ## 🚨 Behavioral Constraints
 
 - **Feature/idea intake**: `/deliver <idea>` (from the `jt-delivery` plugin) — triage → GOAL → author≠verifier orchestration per the `jt-delivery:contract` skill §1a; repo-specific bindings (gates, canon, reviewer routing): `docs/workflows/autonomous-delivery-prompt.md`.
-- **Manager triage (Paul 2026-08-01)**: Paul adds requests mid-flight; triage each against the active plan — now / sequenced / backlog / groom-first — and report the verdict in one line so he can override. Deferring is a decision; interrupting a gated pipeline for a non-critical request is the failure. **Autonomy grant**: handle sprints end-to-end; for big/critical calls run a 2-4 agent voting panel and make the call (split + irreversible → hold for Paul). Record every autonomous call with its evidence. Paul's explicit words override any panel.
+- **Manager triage (Paul 2026-08-01)**: Paul adds requests mid-flight; triage each against the active plan — now / sequenced / backlog / groom-first — and report the verdict in one line so he can override. Deferring is a decision; interrupting a gated pipeline for a non-critical request is the failure. **Autonomy grant**: handle sprints end-to-end; for big/critical calls AND any highly OPINIONATED question (design taste, naming, strategy trade-offs — anywhere reasonable people disagree) run a voting panel of subagents with DISTINCT PERSONAS/lenses and organize an explicit vote (Paul 2026-08-21); make the call (split + irreversible → hold for Paul). Record every autonomous call with its evidence. Paul's explicit words override any panel.
 - **Agent 4-eyes on EVERY stage, author ≠ verifier (Paul 2026-08-20, widened 2026-08-21)**: every change is written by one agent and verified by a different one; review lands BEFORE the artifact leaves the workshop (plan before Paul sees it, diff before commit, finding before reported, measurement before quoted). Brief reviewers with goal + artifact, never conclusions; ask for measurements, not verdicts. Internal sub-agents per stage — **model routing (Paul 2026-08-21): critical decisions, editing/outlining AND design critique on fable; coding, code review AND content writing on opus; mechanical/simple tasks on sonnet; spawns never inherit the session model implicitly — pass `model` on every Agent call**; `/codex:review` once at the merge gate for user-facing changes. Full protocol: `jt-delivery:contract` skill §5; repo routing: `docs/workflows/autonomous-delivery-prompt.md`. (Evidence: 2026-08-20/21 — reviewers caught a CSS-breaking defect, a fabricated quote, and 11 findings on the delivery prompt itself, all after author self-review.)
 - **Async-first (MANDATORY)**: written, discoverable artifacts for every decision, finding, and handoff; sync is the exception, written back same-day. A task is not done until a cold session can read its state with zero questions. Canonical surfaces: `.agents/skills/async-first-communication/SKILL.md`; same SOP in `AGENTS.md`.
 - **Company facts have a canon**: check `.okf/content/claims-canon.md` before publishing any number (founding 2008-09-01; tenure from `site.Params.foundingYear`; "4.8/5 on Clutch" linked, NO review count; founder 22+ vs company 18+). A number with no in-repo source is a defect. Correcting one means sweeping the instruction layer too — `PRODUCT.md`, `docs/business/`, outreach kits have re-introduced bad canon before (4 of 8 published figures were wrong on 2026-08-14).

bin/qtest

@@ -45,6 +45,8 @@ PAGE_TESTS = {
   "free-consultation" => "free_consultation",
   "vibe-code-rescue" => "vibe_code_rescue",
   "friday-report" => "friday_report",
+  # v2 clean-slate rail (ADR-0006): layouts/next/* + pages/next-pilot.css
+  "next-pilot" => "next_pilot",
   "simple-page" => "privacy_policy",
   # critical/privacy-policy-critical.css maps by basename; alias it to the
   # simple-page tests so a touch there doesn't abort as an unknown key
@@ -73,7 +75,10 @@ def pages_for(file)
   when %r{^themes/beaver/assets/css/pages/([^/]+)\.css$} then [Regexp.last_match(1)]
   when %r{^themes/beaver/assets/css/critical/([^/]+)-critical\.css$} then [Regexp.last_match(1)]
   when %r{^themes/beaver/layouts/blog/}, %r{^content/blog/} then %w[blog-list blog-single]
-  when %r{^themes/beaver/assets/css/}, %r{^themes/beaver/layouts/} then :all
+  when %r{^layouts/next/}, %r{^content/next/} then PAGE_TESTS.keys.grep(/^next-/) # every v2 key; shared v2 templates affect them all
+  # Terminal: root layouts/ had NO case at all, so a root-layouts-only diff
+  # mapped to nil and qtest exited 0 having tested nothing (false green).
+  when %r{^themes/beaver/assets/css/}, %r{^themes/beaver/layouts/}, %r{^layouts/} then :all
   end
 end
 

content/next/_index.md

@@ -0,0 +1,10 @@
+---
+title: v2 preview
+# Clean-slate v2 rail (ADR-0006). Everything under /next/ is staging on the
+# production host: never indexed, never listed, never in the sitemap.
+# noindex -> layouts/partials/seo/enhanced-meta-tags.html robots meta
+# private -> layouts/sitemap.xml skips the page
+cascade:
+  noindex: true
+  private: true
+---

content/next/services/fractional-cto.md

@@ -0,0 +1,6 @@
+---
+title: Fractional CTO (v2 preview)
+source_page: /services/fractional-cto
+layout: single
+type: next
+---

docs/adr/0006-clean-slate-dual-run.md

@@ -0,0 +1,139 @@
+# Architecture Decision Record: Clean-slate v2, dual-run with a /next/ staging rail
+
+**ADR-0006**
+**Date:** 2026-08-21
+**Status:** Accepted (approved by Paul in plan mode, 2026-08-21)
+**Supersedes in part:** ADR-0003's *propagate-into-legacy* rollout mechanism,
+for pages not yet shipped. ADR-0003's brand decisions remain in force on the
+live site until the design competition names a winner.
+
+## Title
+
+Stop converting the legacy cascade page by page. Build v2 on a separate rail,
+run both, and cut over one page at a time through a non-indexed `/next/` path.
+
+## Context
+
+ADR-0003 chose a strangler rollout: convert each page bundle in place, one PR
+at a time, propagating the new tokens into the existing FL-derived cascade.
+Five months of evidence say the two halves of that plan performed very
+differently:
+
+| Approach | Outcome |
+|---|---|
+| Born-clean page (blog rebuild, blog-single, course, vibe-code-rescue, friday-report) | shipped; 0–11 `fl-` lines; no reverts |
+| Propagate into legacy (1a.4 footer, the 41-rule eyebrow sweep, #540's first attempt) | reverted, or created the seam it was removing |
+| Buttons (780 call sites) and the 39 remaining blacks | never started |
+
+The pattern is not about skill. A born-clean page owns its whole cascade, so a
+change to it is local and its blast radius is one screenshot key. A
+propagation change lands in a cascade recovered from an obfuscated FL-Builder
+export, where `legacy-theme-skin.css` and a `:not()`-chain anchor rule outrank
+whatever the page declares — so each fix needs a specificity workaround, and
+the workaround is the next defect.
+
+Paul's direction is that the site must read as a top-tier service with a
+top-tier UI. The recovered cascade is not the foundation for that, and the
+cost of making it one is now measurable in reverts.
+
+## Decision
+
+1. **Legacy is frozen.** `themes/beaver/layouts/**` and the existing bundles
+   take defect fixes only. Two ADR-0003 work items are **dropped** as
+   work-done-twice: the 780-site button unification and the 39 remaining
+   hardcoded blacks. They will be resolved by pages moving to v2, not by
+   editing the pages twice.
+2. **v2 gets its own rail**, forked from the theme rather than grown out of
+   it: `layouts/next/baseof.html` + `layouts/next/single.html` +
+   `themes/beaver/assets/css/pages/next-pilot.css`. The rail keeps every
+   FUNCTIONAL part of the theme baseof verbatim (SEO meta, the five schema
+   partials, favicons, analytics — conversion events are attribute-driven, so
+   the partial keeps them firing — service worker, skip link,
+   `<main id="main-content">`) and reuses the chrome partials as they are. It
+   drops the 7 hardcoded `fl-*` body classes, the `.fl-page` wrapper and the
+   legacy slice members (`style.css`, `586.css`, `vendors/base-4.min.css`,
+   `legacy-theme-skin.css`).
+3. **`/next/` is staging on the production host**, not a separate deploy. It
+   is `noindex` (frontmatter cascade → `enhanced-meta-tags.html`) and excluded
+   from the sitemap (`private: true` → `layouts/sitemap.xml`). **No robots.txt
+   disallow** — that would hide the `noindex` meta from the crawlers that need
+   to read it. Staging is **sunset after two clean flips**.
+4. **A staging stub owns only the URL.** Its `source_page` resolves the real
+   page; the v2 layout renders from that single source so copy cannot drift
+   between rails. The canonical points at the source page, emitted from the
+   shared meta partial so exactly one canonical ships.
+   **PR-1's pilot is a RAIL SMOKE PAGE, not a service-page mirror** — measured:
+   47 of the real page's 536 words render, because service copy lives in
+   frontmatter (`overview`, `faqs`) the smoke template does not read. Porting
+   those frontmatter sections is PR-2's FIRST task, before any styling. Do not
+   judge the design against the smoke page.
+5. **Flipping a page keeps the permalink** so no redirects are needed, and the
+   revert is a frontmatter change. But the flip is only safe ONCE the v2
+   template renders everything the page's frontmatter carries (see 4) — setting
+   `type: next` on the real page before that guts it to the smoke-page subset.
+   The flip checklist: v2 template reads every frontmatter section the page
+   uses → baselines for the page key → then `type: next`.
+6. **Homepage carve-out.** A v2 homepage may be built at `/next/` at any time,
+   but the PUBLIC flip honours the ≈2026-09-17 engagement-read gate from
+   ADR-0004 — the homepage is the only surface with enough traffic for the
+   read to mean anything, and flipping it early destroys the baseline.
+
+## Consequences
+
+**The rail is not zero-legacy — navigation.css leaks a body rule.** The v2
+bundle reuses `navigation.css` and `footer.css` for chrome, and
+`navigation.css:20` sets `body { font-family: system-ui; font-size: 18px;
+line-height: 1.6; color: #121212 }` — so v2 pages inherit legacy body sizing
+and a hardcoded colour literal unless the page bundle overrides them. Known,
+accepted for PR-1; PR-2's design pass owns the override (measured: the pilot's
+loose h1 leading is this rule).
+
+### Positive
+
+- Every v2 page is born clean; no page is edited twice.
+- The rail is testable on the production host, on real devices, before any
+  visitor-facing change.
+- Rollback is a frontmatter revert rather than a CSS bisect.
+- Closing a related false-green: root `layouts/` had no `bin/qtest` mapping at
+  all, so a root-layouts-only diff exited 0 having tested nothing. The rail
+  forced the terminal `^layouts/` → escalate-to-all case that fixes it.
+
+### Negative / accepted risk
+
+- **Two rails exist at once.** A chrome partial edited during the dual-run
+  period must be checked on both. Mitigated by the freeze: legacy takes defect
+  fixes only.
+- **If the competition winner deviates from ruby/LIGHT**, this ADR supersedes
+  ADR-0003's brand decisions too, and the already-shipped blog and course
+  surfaces join a re-skin queue. Accepted cost, taken knowingly.
+- **`/next/` is publicly reachable.** Anyone with the URL sees unfinished work.
+  Accepted: `noindex` + sitemap exclusion is the right trade against the cost
+  of a separate deploy target, and the pages carry no claims of their own.
+- The rail duplicates ~70 lines of baseof. Deliberate: a shared baseof with
+  conditionals is exactly the coupling this ADR exists to remove.
+
+### Neutral
+
+- No new dependency, no build-pipeline change. The v2 bundle goes through the
+  same `assets/css-processor.html` as every other bundle.
+
+## Alternatives considered
+
+**Keep propagating (ADR-0003 as written).** Rejected on its own evidence:
+three of three propagation attempts reverted, and the two largest items never
+started.
+
+**A second Hugo site / separate deploy target.** Cleanest isolation, but it
+cannot be reviewed on the real host with the real chrome, and the flip becomes
+a migration rather than two frontmatter lines.
+
+**Redesign in place behind a feature flag.** Same cascade, same specificity
+fights, plus a flag to maintain.
+
+## References
+
+- `docs/projects/2608-site-design-system/20-29-strategy/20.06-clean-slate-v2-plan.md` — the approved plan this ADR records
+- ADR-0003 — the design system whose rollout mechanism this supersedes in part
+- ADR-0004 — why the homepage flip waits on an engagement read
+- `docs/workflows/new-page.md` §v2 rail — which fork a new page belongs on
+- `docs/projects/2509-css-migration/css-bundle-ownership-map.md` — the `next-pilot` bundle row

docs/adr/README.md

@@ -7,3 +7,4 @@
 - `0003-site-design-system.md` — one design system for site chrome ("Rescue Room")
 - `0004-static-site-experimentation.md` — what gets A/B tested, what cannot be, what replaces it
 - `0005-reusable-ai-instruction-layer.md` — extract generic skills/agents/contract as a Claude Code plugin; repos keep thin bindings
+- `0006-clean-slate-dual-run.md` — freeze legacy, build v2 on its own rail, stage at `/next/`; supersedes ADR-0003's propagate-into-legacy mechanism

docs/projects/2509-css-migration/css-bundle-ownership-map.md

@@ -37,6 +37,7 @@ zero FL-Builder export files remain in any slice.
 | not_found | `404.html` | — (404.css, dynamic-404.css) | 16.2K → 16.2K | no-FL |
 | friday-report | `page/friday-report.html` | pages/friday-report.css | new 2026-08-20 | born semantic (no FL modules; one `#fl-main-content` specificity override against legacy-theme-skin) |
 | pagination | `list.html` (second bundle) | — (pagination.css only) | 0.2K → 0.2K | no-FL |
+| next-pilot | `layouts/next/single.html` (repo root, NOT in theme) | pages/next-pilot.css | new 2026-08-21 | v2 clean-slate rail (ADR-0006): css-variables + navigation.css + footer.css only — no style.css/586/base-4/legacy-theme-skin. Serves `/next/*` staging pages |
 
 ¹ careers was a byte-identical verbatim move (3086-layout2 had zero dead
 nodes); the small growth vs 07-12 is upstream content drift (course v2 merge

docs/projects/2608-site-design-system/20-29-strategy/20.07-design-competition-brief.md

@@ -61,9 +61,19 @@ action, not six.
 
 ## Judging (after both directions exist)
 
-- Critic: `/impeccable` critique **on fable** (Paul: design critic = fable),
-  plus ICP cold-eyes panel — lenses: first-5-seconds trust, purchase-intent
-  CTA clarity, mobile.
+- Critic: `/impeccable` critique **on fable** (Paul: design critic = fable).
+- **Persona voting panel (Paul 2026-08-21: highly opinionated questions get
+  multiple subagents with different personas and an organized vote).** Distinct
+  personas, each voting with its strongest objection + evidence:
+  1. *Alex the burned founder* (the ICP, first-5-seconds trust) — sonnet
+  2. *The CFO skeptic* (purchase-intent: would I authorize $15K/mo off this
+     page?) — sonnet
+  3. *The mobile commuter* (390px, thumb-reach, scan-only) — sonnet
+  4. *The brand strategist* (what does this say about us vs every other
+     agency site?) — sonnet
+  Votes are organized into one tally; independent agreement is valid,
+  manufactured dissent is not; each names the strongest finding AGAINST its
+  own preferred direction.
 - Paul sees ONE side-by-side board per direction: homepage + single-service
   screens, 3-line verdict per lens, one "what this says about us" sentence.
 - **Paul picks.** Rejection feedback, if both fail, seeds direction 3.

docs/workflows/new-page.md

@@ -4,6 +4,32 @@ The checklist for building a new page that reuses the shared components and
 stays consistent with the design system. Every trap listed here has bitten a
 real sprint; do not skip steps.
 
+## 0. Pick your fork first — LEGACY or v2 rail
+
+Since ADR-0006 there are **two** paved paths, and §1–8 below describe the
+**LEGACY** one. Legacy is frozen: it takes defect fixes and pages already on
+it, not new design work.
+
+| | LEGACY (§1–8 below) | **v2 rail** |
+|---|---|---|
+| Baseof | `themes/beaver/layouts/baseof.html` (7 `fl-*` body classes, `.fl-page` wrapper) | `layouts/next/baseof.html` (neither) |
+| Shell | `fl-row` / `fl-col-group` / `fl-col` nesting is load-bearing | plain semantic HTML; **no `fl-*` anywhere** |
+| Bundle | `critical/base.css` + `586.css` + `vendors/base-4.min.css` + `style.css` + `legacy-theme-skin.css` + … | `foundations/css-variables.css` + `navigation.css` + `footer.css` + your page CSS. Nothing else. |
+| Where it renders | its own permalink | `/next/<path>/` — `noindex`, out of the sitemap |
+
+**A NEW page, or a page being redesigned, goes on the v2 rail.** Copy
+`layouts/next/single.html` and `themes/beaver/assets/css/pages/next-pilot.css`
+as the starting shape, add a stub under `content/next/` whose `source_page`
+points at the real page (the stub owns only the URL — copy keeps one source of
+truth), and flip the real page with `type: next` ONLY once the v2 template renders every frontmatter section the page uses (ADR-0006 §5 checklist) — flipping before that guts the live page to whatever subset the template reads.
+
+Steps 2 (bundle order), 5 (purge prefix) and 6 (qtest map, screenshot tests,
+ownership map) apply to BOTH forks — a v2 page still needs its `PAGE_TESTS`
+key and its ownership-map row, and `test/unit/paved_path_guard_test.rb` fails
+without them. Steps 1, 3 and 4's `fl-*` guidance apply to legacy only.
+
+Read `docs/adr/0006-clean-slate-dual-run.md` before starting either.
+
 ## 1. Template
 
 Create `themes/beaver/layouts/page/<name>.html`. Start from
@@ -17,10 +43,11 @@ structural shell:
     ... fl-row / fl-col-group / fl-col / fl-col-content nesting ...
 ```
 
-The `fl-*` framework classes are load-bearing (layout grid from
-`critical/fl-layout-grid.css`) — keep the nesting; name YOUR elements with a
-page prefix (`<name>-hero`, `<name>-grid`, ...). Never introduce hash-like
-class names.
+The `fl-*` framework classes are load-bearing **on this fork** (layout grid
+from `critical/fl-layout-grid.css`) — keep the nesting; name YOUR elements
+with a page prefix (`<name>-hero`, `<name>-grid`, ...). Never introduce
+hash-like class names. On the **v2 rail** there is no grid framework and no
+`fl-*` at all — see §0.
 
 ## 2. CSS bundle slice