super-ux 0.41.5 → 0.43.0

package/CHANGELOG.md

@@ -1,5 +1,71 @@
 # Changelog
 
+## 0.43.0 — 2026-08-17
+
+**`ux-flows` has told agents to sweep real products before inventing a flow since 0.35.0,
+and no prompt could ask for it.** The body carries the rule — *"Real flows off the shelf,
+before you invent one"*, with Refero read for structure and Mobbin looked at for whether it
+reads — and `references/funnel-research.md` FR-01 collects live competitor funnels. The
+`description` said flows, wireframes and task analysis, so `найди референсы дизайна`,
+`подбери референсы` and `find reference screens` all reached no route at all.
+
+The word is now advertised: the funnel clause reads *"read against reference screens from
+products already in the category"*, which says the same thing in fewer characters and makes
+`reference screens` / `референсы` routable.
+
+**Why this skill and not the visual one.** `sheleg-design`'s own `DESIGN_SYNC_BRIDGE.md` §4
+opens with *"A reference sweep answers what a good version of this screen contains —
+sections, hierarchy, content order. It never answers what it looks like."* Structure is this
+chain's ground, so the unqualified word lands here; the visual half (`visual reference` /
+`визуальные референсы`) went to `sheleg-design` 1.40.0 in the same pass. A prompt naming both
+— `нужны визуальные референсы` — now raises both routers, which is the wanted answer.
+
+**Budget:** 893 → 940 characters, 30 free under the 970 working limit; the rewritten funnel
+clause paid for most of the addition.
+
+## 0.42.0 — 2026-08-17
+
+**The pack has carried the growth vocabulary since long before it could be asked for it.**
+Counted across `plugins/super-ux`: **448** mentions of funnels, **499** of onboarding, **493**
+of paywalls, **196** of retention, **171** of activation, **100** of referral. `retention`,
+`onboarding`, `paywall`, `winback`, `lifecycle`, `activation`, `virality` and `referral` are
+first-class tags in the practice taxonomy. `references/funnel-research.md` is 190 lines of
+method, `FR-01`..`FR-07`. And not one of those words appeared in any skill's `description`.
+
+That mattered more than it looks, because **a description is a routing surface**. The
+umbrella's prompt hook may only fire on words the skill itself advertises — the fixture that
+enforces it does a literal substring check — so fifteen of fifteen realistic growth prompts
+reached no route at all: `как повысить ретеншн пользователей`, `спроектируй воронку
+активации`, `добавь пейволл после онбординга`, `improve user retention`, `reduce churn on the
+trial`, `add a paywall screen`. Measured by `test/route_coverage.js` in the umbrella.
+
+### Two descriptions, and which word goes where is not a preference
+
+`funnel-research.md`'s own `FR-07` already says where each finding lands, so the split follows
+it rather than convenience:
+
+| Word | Skill | Why that one |
+|---|---|---|
+| funnel, onboarding, paywall, activation funnel | `ux-flows` | *"The funnel's step chain and its branches → `docs/ux/flows.md`"* — and this skill's body already names the chain (ad → landing → quiz → loading → offer → paywall → checkout → success) |
+| user retention, churn | `ux-foundation` | *"Who the buyer is, what job they are hiring the product for → `foundation.md`"*; a journey covers before, **during and after**, which is where retention is won and churn starts |
+
+Neither skill could carry the other's words honestly, which is why the route now fronts two
+skills rather than one.
+
+### The English half is phrases and the Russian half is bare words, and that was measured
+
+A first pass advertised the bare `activation` and `retention`. The umbrella's matcher stems
+them to `activat-` and `retent-`, and the probe caught the cost immediately: `activate the
+virtualenv` and `activate the feature flag` routed to `/ux`, and so did `retention policy for
+logs`. The English words became `activation funnel` and `user retention`; «активация» and
+«ретеншн» stayed single words, because nobody here writes «активация» about a virtualenv.
+After the change all three noise cases are silent and every growth prompt still routes.
+
+**Budget:** `ux-flows` 592 → 884 characters, `ux-foundation` 401 → 619. Both inside the 970
+working limit with room left, which was the point of not spending it on the bare words.
+
+`python3 test/validate.py` → `OK (3500 checks)`; the full gate green.
+
 ## 0.41.5 — 2026-08-16
 
 **`copywriting` now answers when somebody asks for a landing page.** The unqualified
@@ -90,6 +156,11 @@ Both are recoveries from the same mistake, and only one of them is safe.
 
 ## 0.41.2 — 2026-08-16
 
+> **Never released on its own.** There is no `v0.41.2` tag and no `0.41.2` on npm,
+> so `npm install super-ux@0.41.2` and `git checkout v0.41.2` both fail. This section
+> describes work that shipped inside a later version. The note is here because
+> the section reads as a release (2026-08-17, umbrella `B-71`).
+
 **A source file is not prose, and the prose rules were reading all of it.** On
 `sshlg.me`, whose `Sources:` block points `marketing` at `src/data/*.ts`, that
 produced 20 rhetorical-dash errors inside `//` comments and 7 keyword-stuffing
@@ -144,6 +215,11 @@ buried under twenty false ones.
 
 ## 0.41.1 — 2026-08-16
 
+> **Never released on its own.** There is no `v0.41.1` tag and no `0.41.1` on npm,
+> so `npm install super-ux@0.41.1` and `git checkout v0.41.1` both fail. This section
+> describes work that shipped inside a later version. The note is here because
+> the section reads as a release (2026-08-17, umbrella `B-71`).
+
 **`B024` fined the writer for meeting a threshold this pack sets.** The
 sentence-case check reads any capitalised word inside a sentence as Title Case,
 and a contraction of the first person survived every exemption it had: `I'm` is
@@ -1247,6 +1323,11 @@ copy.
 
 ## 0.25.0 — 2026-07-29
 
+> **Never released on its own.** There is no `v0.25.0` tag and no `0.25.0` on npm,
+> so `npm install super-ux@0.25.0` and `git checkout v0.25.0` both fail. This section
+> describes work that shipped inside a later version. The note is here because
+> the section reads as a release (2026-08-17, umbrella `B-71`).
+
 The catalog knew how to sell inside an app and almost nothing about selling
 on the web. Two funnels were missing: web-to-web (landing → pricing →
 checkout → recurring billing → cancel) had only generic form advice, and

package/package.json

@@ -1,6 +1,9 @@
 {
   "name": "super-ux",
-  "version": "0.41.5",
+  "version": "0.43.0",
+  "scripts": {
+    "test": "python3 test/validate.py && python3 test/brand_lint_test.py && python3 test/ux_lint_test.py && python3 docs/ux/lint.py && python3 docs/brand/lint.py"
+  },
   "description": "Scenario-driven UI development for AI agents (Claude Code, Cursor, 70+ agents): a versioned design chain in docs/ux/, a scenario-first hard rule, a deterministic drift linter, and evidence-backed UX audits. This package is the installer CLI.",
   "bin": {
     "super-ux": "bin/super-ux.js"