@am_shork/attest 1.1.0 → 1.1.1

package/CHANGELOG.md

@@ -90,6 +90,149 @@ above and 110 stale lines on this one.
 security hole takes a major, on whatever schedule a hole is found. That price is
 now stated rather than unknown, which is what made the line decidable at all.
 
+## [1.1.1] - 2026-09-03
+
+### Documentation
+
+- **`docs/{en,zh}/intent-quality.md` gains a seventh finding: a param no
+  statement interpolates changes with nothing outside the line to show it.**
+  `render --check` gives a review surface for the 92.1% of this repository's
+  params that a statement does interpolate — editing one moves `SPEC.md`. It
+  cannot help with the rest: five are structurally uninterpolatable (a
+  kind-to-payload table or a list, which `non-scalar-interpolation` refuses in
+  a statement) and two are scalars nobody interpolated. For those, every gate
+  treats the edit as fine — `check` sees a schema-valid literal,
+  `possible-drift` is satisfied by any covering scenario reading the param at
+  all, and `verify` stays green wherever the assertion derives its expectation
+  from the same value — so a human reading the registry diff is the only thing
+  that catches an unintended move. Split out of the *Under consideration* entry
+  on changing an existing param's value, as the cheap half it named while the
+  WARNING it also considered stays undecided for want of a second corpus.
+
+### Added
+
+- **`pre-commit-check` gained §9, a hand sweep for a dependency bump.** Closes
+  the *Considered and rejected* entry on a standing `no-deprecated` gate: three
+  deprecated call sites across a Zod 3→4, Vitest 3→4 and typescript-eslint
+  8.2→8.65 walk was too small a denominator to carry a third tsconfig or an
+  `allowJs` admission into the wider typed project, so the check stays a manual
+  step — run `@typescript-eslint/no-deprecated` against a throwaway project that
+  also covers the five files no tsconfig in this tree includes
+  (`eslint.config.js`, `bin/attest.js`, the three `vitest*.config.ts`) — rather
+  than a permanent gate.
+
+- **A gate on which files inside this tree may name the package by the name it
+  publishes under.** A built `dist/` makes this package resolve *itself*:
+  `package.json` names the subpaths in `exports`, so `@am_shork/attest/define`
+  resolves to `dist/core/registry.js` on any machine that has run a build. CI's
+  lint, test and typecheck jobs run before one, and there it resolves nowhere.
+  Measured once, on an archived `requirements.delta.ts` written with the
+  specifier the workflow document hands *adopters*: it passed `eslint`, both
+  typechecks, the whole suite, `check self`, `verify self` and `archive --apply`
+  locally — **six gates** — then failed **three CI jobs** on that one line. This
+  is the sibling of the `bin/attest.js` loads `dist/` gotcha and the more
+  dangerous half of it: there a stale build runs old code, here the build
+  vouches for an import that has no other resolution.
+
+  `tests/import-boundary.spec.ts` now reads the whole checked tree through the
+  same AST walk the runner-boundary tests use, and asserts the set of published
+  specifiers against a one-entry allowlist. The single legitimate namer is
+  `examples/judgement-table/judgement.spec.ts` at `@am_shork/attest/inspect`,
+  and it is legitimate for exactly the reason the rule exists: `./inspect` is
+  the only subpath with somewhere to resolve *to* inside this tree, via
+  `resolve.alias` in `vitest.config.ts` and the matching `paths` entry in
+  `tsconfig.typecheck.json`.
+
+  *Two things about the walk are load-bearing and neither is obvious.* It does
+  not use `findFiles`, because `SKIP_DIRS` excludes `changes/` and `archive/` —
+  correctly, for a normal scan — while `tsc` and `eslint` are handed `self/**`
+  and walk straight into both, and the import this exists for was in an
+  *archived* delta; the engine's own scanner would have produced a gate that
+  passes over the only file it was written about, which is what the first draft
+  of it did. And it covers the repo root as well as the four directories
+  `tsconfig.typecheck.json` includes, because `vitest.config.ts`,
+  `vitest.self.config.ts` and `vitest.consumer.config.ts` are named by no
+  `include` in the tree yet Vitest resolves them **first**, before it can
+  collect a single test — a published specifier there fails CI earlier and
+  harder than one under `self/`, and the directory-shaped version of this gate
+  would not have looked. Each half is pinned by an instrument guard, so a walk
+  that quietly stops reading `examples/` or the root fails on the guard rather
+  than passing empty.
+
+- **A product identity, as delivery artifacts.** `assets/brand/` carries the
+  mark, four icon sizes and four lockups as SVG, with `assets/brand/README.md`
+  as the guide: the geometry grid, the mortise-and-tenon overlap that must not
+  become clearance, the scaling rules, the type stack and the palette. The
+  README header and `package.json`'s keywords are the rest of it.
+
+  **Nothing in `src/` reads any of it, and that is the shape of the change
+  rather than a stage it has not reached yet.** A CLI half was built and dropped
+  in the same session — a palette module plus a `--help` banner — and the
+  argument that closed it is filed under *Considered and rejected*.
+
+### Changed
+
+- **Three deprecated call sites retired, found by turning
+  `@typescript-eslint/no-deprecated` on the whole typed tree rather than by
+  reading the diff.** Neither was reachable by the repo's standing gates: both
+  compiled, linted and ran green, because a deprecation is a JSDoc tag and
+  nothing in `eslint.config.js` was asking for it.
+
+  `src/core/schema.ts` passed `{ message: … }` to the `statement` refinement.
+  Zod 4 deprecated that parameter in favour of `error`, and the rest of the file
+  had already moved — `paramValue`, `RegistryRecord` and `RegistrySchema` all
+  spell it `error` — so this was the one survivor of an incomplete migration,
+  not a considered exception. The wording is unchanged, and by the
+  `Issue.message` rule it is prose either way, so nothing observable moved:
+  parsed against the built schema, the refusal still comes back with the same
+  `message` under the same `custom` code. Design §5.1 quotes this schema in both
+  languages and carries the same edit.
+
+  `tests/registry-types.spec-d.ts` used `expectTypeOf(…).toMatchTypeOf<…>()`,
+  deprecated in expect-type 1.2.0 and split into two successors. Both sites
+  assert *assignability to a supertype* — `Registry` for the whole registry,
+  `Requirement` for one entry — which is `toExtend`, the successor that keeps
+  the `extends` relation. `toMatchObjectType` is the other one and would have
+  been the wrong reading: it checks a strict subset of keys, which is a
+  different claim than the one these two tests were making.
+
+  `eslint.config.js` called `tseslint.config()`, deprecated in
+  typescript-eslint 8.65 now that ESLint core provides the same `extends`
+  flattening via `defineConfig()`. Swapped for `defineConfig` + `globalIgnores`
+  from `eslint/config`. Because both spellings lint clean, *clean was not
+  evidence*: equivalence was established by diffing `eslint --print-config` for
+  a file in each of the five trees across the swap — byte-identical — plus the
+  linted file set and problem count, 166 and 0 on both sides.
+
+  **The third one is the interesting one, because the first sweep structurally
+  could not see it.** `no-deprecated` is type-aware, so it only reaches files
+  some project includes — and `tsconfig.typecheck.json` includes `src`, `tests`,
+  `self` and `examples`, which leaves `eslint.config.js`, `bin/attest.js` and
+  the three `vitest*.config.ts` named by no `include` in the tree. This is the
+  same blind spot `tests/import-boundary.spec.ts` was written about from the
+  other direction: the root config files are read *first* by the tooling and
+  belong to no project last. Re-run against a throwaway project that includes
+  them, the remaining four are clean — and that result was itself checked
+  against the failure shape this repo keeps meeting, by re-running the scan on
+  the pre-fix `eslint.config.js` and confirming it goes red, since a scan
+  matching no files is also silent.
+
+  With that gap closed — and the installed dependency tree carrying no registry
+  deprecation, the adopter-facing samples inside `skill.ts` and every code fence
+  in `docs/` and the README checked against the deprecated identifiers harvested
+  from the dependency typings, `fixtures/**` scanned under a throwaway project,
+  and the suite run under `--pending-deprecation` surfacing no runtime `DEP`
+  warning — this is the whole inventory as of Zod 4.4, expect-type 1.4,
+  typescript-eslint 8.65, Vitest 4.1, TypeScript 5.9 and Node 24.
+
+  **Two halves of that are unmeasured rather than clean, and are named here
+  rather than left to be inferred from the version list.** `package.json`
+  declares `typescript` as `^5.5.0 || ^6.0.0` while 5.9.3 is what is installed,
+  so anything TypeScript 6 deprecates has not been looked at; and `engines`
+  allows Node 20.19 while the sweep ran on 24, so a `DEP` warning that fires
+  only on an older runtime would not have appeared. Nothing gates any of it
+  going forward — see *Under consideration*.
+
 ## [1.1.0] - 2026-09-02
 
 ### Added
@@ -7098,6 +7241,15 @@ have to gather again.
   `intent-quality.md`, the four things no gate is looking at already being that
   page's subject and this being a fifth of the same kind.
 
+  **The cheap half is written under `[Unreleased]`, and the decision above is
+  unchanged by it.** `docs/{en,zh}/intent-quality.md` now carries a seventh
+  finding: a param no statement interpolates changes with nothing outside the
+  line to show it, so the discipline that catches an unintended move there is a
+  human reading the registry diff. That costs no engine surface and settles
+  nothing this entry is still waiting on — the WARNING in blocker **(2)** is
+  still a rule with no measured population, and this is the fallback the entry
+  itself named for exactly that state, not a smaller version of the rule.
+
 - **The intent layer is separable from the repository it describes, and the
   verification layer is not — so a third party can audit intent and never
   obtain a verdict.** Filed 2026-08-22 from a design evaluation of using Attest
@@ -7399,47 +7551,6 @@ have to gather again.
   cost somebody a round trip. Without the first, this trades a derivable field for
   an underivable duplicate.
 
-- **A Stop hook that will not let an agent end a turn while `check self` is red.**
-  From the same read: `nostics` ships `hooks/consume-diagnostic.mjs`, a Claude
-  Code Stop hook that reads the log its dev-server collector writes and returns
-  `{"decision": "block"}` with the diagnostic as the reason, so the agent keeps
-  working rather than stopping on a red application. Three details there are the
-  design and not the plumbing, and are the transferable part: it guards on
-  `stop_hook_active` so a blocked stop cannot loop; it feeds back **one**
-  diagnostic per turn rather than the whole log; and when it consumes one it drops
-  every other entry sharing that code and file, so a fault at twenty call sites is
-  one round trip and not twenty.
-  *Why the shape fits this repository specifically.* `CLAUDE.md`'s "Before
-  committing" and the whole of `pre-commit-check` are discipline: nothing runs
-  `check self` unless a person or an agent remembers to, and the repeat-grouping
-  that hook does by hand is something `formatIssues` already computes. `.claude/`
-  here holds five skills, no hooks and no `settings.json`, so this would be the
-  first of its kind rather than an extension of anything.
-  *Measured cost, 2026-09-01.* `node bin/attest.js check self --json` runs in
-  **1.04–1.57 s** over three consecutive runs; `status self --json` in 0.78 s.
-  That is paid at the end of every turn, which is the number to weigh rather than
-  the wall clock of a single invocation.
-  *What blocks it is the stale-`dist/` trap, and a hook is the worst place to meet
-  it.* `bin/attest.js` loads `dist/`, not `src/` — the Gotcha this repository
-  already keeps — so a Stop hook reports on the last build. After a `src/` edit it
-  blocks on a diagnostic the working tree has already fixed, or clears while the
-  tree is broken, and in both directions it is a gate arguing about a state nobody
-  is in. Either the hook builds first, adding seconds to every turn, or it runs the
-  engine from source and stops being the thing an adopter could copy.
-  *And the larger question it sits in front of.* If this is right here it is right
-  for an adopter, which makes it a question about what `attest init` writes rather
-  than about `.claude/` in this repository. A hook that can block an agent's stop
-  is a much larger commitment than a document, and the rejection quoted in full
-  inside the `init --check` entry in this section — the gate on the installed
-  skill's freshness — turned on exactly the distinction it needs:
-  a signal about work the reader actually did, against a chore the tool imposes. A
-  block on `check` is the first kind, which is the argument for it, and that
-  argument should be made explicitly rather than arrived at by shipping.
-  *What would move it.* Running it here, unshipped and repo-local, for a few weeks,
-  and counting how often it blocks on something that would otherwise have been
-  committed. If that count is zero the discipline was already working and the hook
-  is overhead; if it is not, the number is the case for `init` writing one.
-
 - **An outward-facing form of the judgement tables, so an adopter can run this
   counter-pressure on their own registry.**
   `tests/intent-rule-candidates.spec.ts` maintains four hand-judged tables over
@@ -7675,6 +7786,68 @@ it sat between 0.2.0 and 0.1.7 for two releases, where standing still meant
 sinking one version deeper each time a release was cut above it, and a rejection
 filed under a version reads as belonging to it.
 
+### A brand palette and banner in the CLI
+
+Built, measured green on every gate, and dropped 2026-09-03 in the session that
+built it. Filed here rather than deleted, because the half that was wrong is not
+the half that looks wrong, and the next person to propose "let's put the brand
+colours in the CLI" should reach this before rebuilding it.
+
+**What it was.** `src/cli/brand.ts`: one module holding the palette behind a
+semantic interface — `pass`, `drift`, `muted`, `note`, `id`, plus the three
+marks — with `cli/report.ts` and `cli/index.ts` rewritten onto it so `chalk` was
+imported in exactly one place. Attest Green for a passing verdict, Drift Orange
+for a WARNING, Slate for every secondary line, and a four-row ASCII mark above
+`--help` on a TTY.
+
+**The deletion test passed, which is why this is worth recording.** Delete the
+module and the same seven colour decisions reappear across two files and thirty
+call sites, spelled slightly differently each time — that is not a pass-through,
+and it is the ordinary reason this repository consolidates something. The gates
+agreed: lint, both typechecks, the full suite, `check self`, `verify self` and
+`render --check` were all green, and `attest --help` through a pipe was
+byte-identical, so `self/cli-surface.spec.ts` saw nothing.
+
+**What it could not answer is who the interface is for.** A module earns its
+keep by having callers who vary across it, and this one had two callers that
+never will: there is no second theme, no `--color` flag, no second renderer, and
+nothing in the roadmap proposes one. *One adapter is a hypothetical seam.* The
+identity was consolidated at a place where nothing varies, which buys a
+re-tunability nobody has asked to use, and every colour in it is invisible to
+every test in the tree — they strip SGR before asserting, correctly, because
+`Issue.message` is prose and so is its colour. So the module's whole surface is
+unassertable: **the interface is the test surface, and this interface had none.**
+
+**Three findings survive the rejection, and they are the reason it is written
+out.** They cost a session to learn and none of them is recoverable from the
+diff, which no longer exists.
+
+- *Frost (`#E6EAF2`) can never be written by a CLI.* The guide defines it as
+  "text on Void", and a terminal's background belongs to the reader — Frost on a
+  light profile is invisible. The terminal's own foreground already plays that
+  role. The same argument retires Void, Surface and Paper: a CLI that paints its
+  own ground fights the profile the reader picked.
+- *Drift Orange cannot become the ERROR colour.* The palette has no failure
+  colour, and the guide makes Drift Orange's scarcity part of what the product
+  means — it appears when something is wrong. Spending it on every ERROR spends
+  the meaning. The terminal's red is not a placeholder for a brand colour; it is
+  the right answer.
+- *A monospace mark must be ASCII, and the reason is width, not encoding.*
+  `█`, `╱`, `╲` and `═` are all East Asian **ambiguous** width, so a CJK
+  terminal — which is where this repository is read — renders them two cells
+  wide, and mixing them with a one-cell `=` or space tears a multi-row figure
+  apart. The CLI's existing `✓`, `✗` and `●` are ambiguous too and survive it
+  only because each is alone on its line with nothing aligned under it. This
+  generalises past branding: it applies to any table, tree or rule the CLI ever
+  draws.
+
+*One thing found on the way is not a branding decision and does not belong to
+this rejection*: `chalk.dim` is SGR 2, which some profiles render at 40% opacity
+and others ignore outright, so the same secondary line is either unreadable or
+undifferentiated depending on the terminal. That is a legibility defect in
+output that ships today, and it survives this entry as a candidate on its own
+terms rather than as a reason to bring the palette back.
+
 ### A backlink index over `archive/`, recording which change introduced or renamed an id
 
 Proposed from OpenViking's `.relations.json`, restated twice on measurement, and
@@ -9407,6 +9580,47 @@ declared prefix ownership recorded above in this section — rejected on there
 being no adoption that needs it, which is the same condition this sentence
 supposes — which addresses, rather than a scorer, which ranks.
 
+### A standing gate on deprecated API usage
+
+Priced on 2026-09-02, on the sweep that retired three deprecated call sites
+under `[Unreleased]` — turning `@typescript-eslint/no-deprecated` on by hand
+over a dependency walk through Zod 3→4, Vitest 3→4 and typescript-eslint
+8.2→8.65, none of the three reachable by anything the repo runs on a commit.
+The obvious move was to promote the rule into `eslint.config.js` beside
+`no-unnecessary-condition`, itself promoted from `strictTypeChecked` on exactly
+this argument: it needs the checker, and it found real instances here.
+
+*What blocked it was not the rule, it was the reach.* `no-deprecated` is
+type-aware, so it only sees files a project includes, and the one deprecated
+site with the widest blast radius — `tseslint.config()` in `eslint.config.js`
+itself — sits in the set no project includes, along with `bin/attest.js` and
+the three `vitest*.config.ts`. Turning the rule on as-is buys a gate that
+covers `src`, `tests`, `self` and `examples` and silently *not* the files the
+tooling reads first, which is the shape this repo has been burned by twice
+already: `tests/import-boundary.spec.ts` deliberately does not use `findFiles`
+for the same reason, and its instrument guards exist because a walk that
+quietly stops reading a tree passes empty. A gate whose coverage is
+believed-wider-than-it-is is the failure this file keeps recording. Closing the
+gap meant either growing `tsconfig.typecheck.json` over the root config files —
+pulling `allowJs` into the wider project for `eslint.config.js` and
+`bin/attest.js` — or a second small project just for them, a third tsconfig in
+a repo that deliberately has two and has a comment in each explaining why.
+
+**Rejected on the denominator.** Three call sites across an entire dependency
+walk is too small to carry the cost of either tsconfig shape — permanent, on
+every file the wider project would newly admit — against a defect a hand sweep
+at upgrade time already catches. `pre-commit-check` §9 records that sweep: run
+`no-deprecated` against a throwaway project covering what
+`tsconfig.typecheck.json` does and the five files it does not, and confirm the
+scan is actually seeing them before trusting a clean result. The check still
+happens, on the cadence the defect actually arrives on, without a third
+tsconfig or `allowJs` admitted into a wider project.
+
+*Worth reopening if the count moves* — a dependency bump that turns up more
+than a handful of deprecated sites in one sweep is the signal that a hand pass
+is no longer keeping up, and it is measured the same way this entry was: read
+the `no-deprecated` output before deciding whether it needs a permanent home.
+
 ## How an entry is re-measured
 
 What the three sections above have learned about themselves, kept apart from all

package/README.md

@@ -1,4 +1,9 @@
-# Attest
+<p align="center">
+  <picture>
+    <source media="(prefers-color-scheme: dark)" srcset="https://gitlab.com/Pseudorca/attest/-/raw/main/assets/brand/attest-logo-horizontal.svg">
+    <img src="https://gitlab.com/Pseudorca/attest/-/raw/main/assets/brand/attest-logo-horizontal-light.svg" alt="Attest" width="260">
+  </picture>
+</p>
 
 A TDD-native spec framework. **Tests are the source of truth for verification;
 ID-bound requirements are the source of truth for intent.** Attest binds the two
@@ -159,7 +164,7 @@ Every diagnostic carries a `code`, and every code has a section in
 ```
 ERROR   registry-not-static (requirements/upload.reqs.ts:5)
     Value is not a literal.
-    → https://gitlab.com/Pseudorca/attest/-/blob/v1.1.0/docs/en/troubleshooting.md#registry-not-static
+    → https://gitlab.com/Pseudorca/attest/-/blob/v1.1.1/docs/en/troubleshooting.md#registry-not-static
 ```
 
 The anchor **is** the code, so the link cannot point somewhere the section

package/dist/core/schema.js

@@ -42,7 +42,7 @@ export const RequirementSchema = z.object({
     statement: z
         .string()
         .refine((s) => /\b(SHALL|MUST)\b/.test(s), {
-        message: 'statement must contain the RFC-2119 keyword SHALL or MUST',
+        error: 'statement must contain the RFC-2119 keyword SHALL or MUST',
     }),
     rationale: z.string().min(10, 'rationale must not be empty (the intent layer has to say why)'),
     // A param is any JSON value. Lists (vendor blacklists, id sets) are the most

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@am_shork/attest",
-  "version": "1.1.0",
+  "version": "1.1.1",
   "description": "TDD-native spec framework: tests are the source of truth for verification, ID-bound requirements the source of truth for intent.",
   "type": "module",
   "packageManager": "pnpm@10.28.0",
@@ -64,9 +64,12 @@
   "keywords": [
     "tdd",
     "spec",
-    "requirements",
     "drift",
-    "vitest"
+    "requirements",
+    "traceability",
+    "requirements-traceability",
+    "vitest",
+    "typescript"
   ],
   "license": "MIT",
   "dependencies": {