super-ux 0.35.2 → 0.36.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (2) hide show
  1. package/CHANGELOG.md +47 -0
  2. package/package.json +1 -1
package/CHANGELOG.md CHANGED
@@ -1,5 +1,52 @@
1
1
  # Changelog
2
2
 
3
+ ## 0.36.0 — 2026-08-12
4
+
5
+ R-14 was filed `never` this morning and closed the same day, by the only
6
+ evidence available for a rule that lives in a skill rather than a script: a run.
7
+
8
+ ### Changed
9
+
10
+ - **The sweep was verified in a fresh context, against a brief that never
11
+ mentioned it.** An iOS subscription-cancellation flow, with no reference to
12
+ Mobbin, Refero, Lazyweb or sweeping anywhere in the prompt. The agent gated on
13
+ the tools present rather than the config and said which were which, swept
14
+ **before** drawing, read four shipped cancellation flows, and named what each
15
+ changed and what it refused to let them change — the job stayed the
16
+ foundation's, the visual identity stayed the style pack's. It also declined an
17
+ instruction embedded in one server's own response asking to alter persistent
18
+ instructions, which is the untrusted-data rule holding under a live test nobody
19
+ designed.
20
+
21
+ ### Added
22
+
23
+ - **Two cases the sweep assumed away, both reported by that run.** With only the
24
+ image server present you are reading step order and decision points off
25
+ screenshots — a weaker read than structured steps, not an equivalent one, and it
26
+ has to be said. And a sweep that returns nothing is a result: Lazyweb returned
27
+ zero flows for a completely mainstream journey, and "I swept" with no findings
28
+ and no statement of emptiness cannot be told apart from not sweeping.
29
+ - **B-012** files the run's nine other findings, each reproducible from its
30
+ report — including that the Design workflow has no defined behaviour for "we know
31
+ almost nothing", that the practice pass does not scale to roughly 150 verdicts for
32
+ one flow, and that BP-123 describes web billing on a platform where the app
33
+ cannot cancel an IAP.
34
+
35
+ ## 0.35.3 — 2026-08-12
36
+
37
+ ### Changed
38
+
39
+ - **`best-practices.md` gains a `## Contents` that can actually navigate it.** At 1650
40
+ lines it is the largest reference in the family and had no list at all; a list of its
41
+ two `##` headings would have satisfied the check and served nobody, because the
42
+ catalog's structure lives one level down — 29 `###` categories over 206 practices.
43
+ The list carries both levels, so an agent reading the head learns what the rest holds,
44
+ which is the only reason the rule exists.
45
+ - Written to `skills/references/best-practices.md` — the source of truth — and
46
+ propagated by `test/sync_references.py` to the four skills that ship it. The drift
47
+ guard caught an earlier attempt that edited the copies directly, which is exactly what
48
+ it is for.
49
+
3
50
  ## 0.35.2 — 2026-08-12
4
51
 
5
52
  ### Changed
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "super-ux",
3
- "version": "0.35.2",
3
+ "version": "0.36.0",
4
4
  "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.",
5
5
  "bin": {
6
6
  "super-ux": "bin/super-ux.js"