Both repeat a habit this README already documents, so they belong beside it
rather than as new entries.
`git grep -c` counts lines: a symbol that a file imports and then calls is two
lines, never one. And a check must not grep for a word that the ticket's own
mandated snippet contains — RD-23 required the upload controller's
`dispatch:` property and then asserted zero occurrences of `dispatch`.
The agent hit both, refused to distort the code, and corrected the ticket text.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Move the adres, beroep and controle cases out of registratie-wizard.component.ts
into adres.step.ts, beroep.step.ts and controle.step.ts, matching RD-22's
*.step.ts convention. The parent drops from ~568 to 274 lines and loses its
`eslint-disable max-lines`.
The upload controller moves into beroep.step.ts and emits `uploadMsg` instead of
dispatching directly; the parent maps that back onto the machine's `Upload`
message. `onDiplomaKeuze` stays in the parent (message construction from the DUO
payload belongs in the container) and now takes only the chosen id, reading its
own `duoData` computed instead of receiving the DUO payload as an argument. Each
step injects `RegistratieLookupStore` directly for its own async presentation
(adresStatus, the DUO lookup, samenvattingVragen) — the sanctioned exception,
since it is a root singleton. Markup moved verbatim; the `@@` id count across
the directory stays 43.
Two of the ticket's acceptance numbers do not hold against correct code and are
corrected in the ticket file: `createUploadController` is 2 lines (import +
call), not 1 — `git grep -c` counts lines, and the same shape gives 2 for
`createStore` and 3 for `createDraftSync` elsewhere in this codebase. `dispatch`
is 1, not 0 — decision 4's mandated `UploadControllerDeps.dispatch` property
name is that string even though it is not the machine's dispatch.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
The parent held one @switch with three @case blocks — three screens'
markup in one file. Each case is independent and needs only the
answers, the errors, and (for two of them) the scholing threshold.
Extract buitenland.step.ts, werk.step.ts, and review.step.ts as pure,
presentational steps: inputs down, one narrow output up, dispatch
never passed down. The parent keeps the store, the shell, and
draftSync, and maps each step's output back to a machine message.
This is the first *.step.ts in the repo, so it sets the naming
convention that RD-23 does the same job with.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
deleteAdjacentChip and insert did getSelection()/Range work inside the
component, which pushed it over the max-lines budget under a disable
comment. rich-text-dom.ts already owns the DOM boundary, so the surgery
moves there as two new exports, chipAtCaret and insertChipAtCaret, and
the component keeps only its event-handling and output concerns.
adjacentChip stays exported with its own spec case. The component
disable comment is gone, since the file is now under the line budget.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
RD-03 moved the dashboard page into the `overzicht` context and RD-04 left the
folder name alone, recording it as stale. It stays stale: the folder is named
after a page that lives somewhere else, so a reader looking for the dashboard
finds four sections that are not it.
Measured cost: 8 `git mv`s and four import lines in one file. Fold in two doc
paths that still name the moved page.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
RD-20 asserted `git grep -l "toWizardErrors"` would find 5 files, but its own
Steps list regenerates `behaviour-spec.mdx`, and the generator publishes every
`describe` title. Naming a spec after the function it tests puts the name in
the generated document too, so the honest count is 6.
The agent refused to rename the describe block to satisfy the number, which is
the correct response and matches the precedent from RD-14.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Each wizard flattened its per-field error record into the shell's
WizardError[] summary with its own copy of the same loop. Extract one
pure helper, wizard-errors.ts, next to naarStapLabel. Add a spec that
covers a flat record, an empty record, skipped undefined/empty-string
values, the idPrefix, and a skipped nested object.
registratie-wizard.machine.ts changes Errors from an interface to a
type alias, because only a type alias gets an implicit index
signature and is assignable to the helper's Record<string, unknown>
parameter. The other two machines already declare their error maps as
type aliases, so this also makes the three consistent.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
RD-18 established that a ticket reference inside a path to a document that
still exists is a pointer, not provenance, and exempted one. RD-19 did not
re-run that check against `backend/`, and asserted a target of 0. Two such
paths existed, so the agent described the two documents in prose instead. The
path no longer resolves and the reader must search.
Restore both paths, correct the ticket's decision 1 and acceptance target to
2, and record the miss as the eighth in the README's list.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The backend half of the sweep RD-18 did for the front end. git blame
holds the provenance and stays correct when the code moves; the
comment names a closed ticket and tells the reader nothing the
sentence around it does not.
public/letter.css and LetterHtml.golden.html change together, because
the renderer inlines the CSS and the golden file snapshots the
result.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
204 WP-NN/RB-NN comments named a closed ticket instead of the code they
sit next to. git blame already records history and stays correct when
code moves; the comment does not. This sweep removes the reference and
keeps the sentence, across 95 files in apps/ and libs/ plus the
behaviour-spec generator's header text.
Eleven references stay: five story files justify an a11y disable per
the README's rule, and one line in a11y.mdx documents that convention.
Two sentences needed a rewrite, not a deletion, so the reference's
meaning survives its removal. behaviour-spec.mdx is regenerated, not
hand-edited.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Three dashboard-section stories used a fourth title segment
(`Domein/Registratie/Dashboard/<Name>`) that the sidebar rule does not have.
Drop the `Dashboard/` segment so all story titles follow the one rule from
CLAUDE.md decision 5.
Add the two missing stories for sections that have more than one visual
state: `wat-moet-ik-regelen` (4 states) and `wat-wilt-u-doen` (2 states, the
first story in the new `Domein/Overzicht/` bucket). `beheer-links` gets no
story — it has one visual state and its other branch renders nothing.
The `MetTaken` story disables the a11y addon with the same reason and WP-11
reference already used on `task-list.stories.ts` and `choice-list.stories.ts`:
`app-choice-link`'s host sits between the keuzelijst `<ul>` and its `<li>`,
a pre-existing structural gap this ticket does not fix.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
`/clear` ends the supervisor's context. It does not stop the developer agent
that session spawned. A fresh supervisor read the half-finished tree as an
interrupted session and dispatched a second agent onto the same ticket. Both
agents then ran `dotnet test` in one checkout, which truncated the shared
SQLite file and produced the stale-database failure the Troubleshooting
section already records under a different cause.
Also record that a bare `dotnet test` runs the OpenZaak integration test that
both CI scripts filter out.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Eight sites hand-rolled `rd.tag === 'Success' ? rd.value : fallback`. Six
take the new `successOr(rd, fallback)`, one takes the existing `successOf`,
and one (`big-profile.store.ts`) uses the existing `map`, since it returns a
RemoteData rather than an unwrapped value.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
RD-16 was to make parseDashboardView return BigProfile directly, on the
plan's claim that DashboardViewDto, DashboardView and BigProfile were
three names for one payload. Reading the type disproves it.
DashboardView is a pair of BigProfile and HerregistratieDecisions.
BigProfile is { registration, person } and has nowhere to put decisions,
so returning it directly would silently drop the server-computed
herregistratie eligibility — the value ADR-0001 says the front end must
render rather than recompute.
The store's two map calls are not a redundant hop either. They project
one aggregate into two independently consumed signals, and six files read
them separately.
Also withdraw the earlier correction that "Step 2 did not fully land".
That claim came from reading the parse signature without reading the type
it returns. Commit 42e7a1e did the right parts, including moving
HerregistratieDecisions into domain, and correctly left alone the part
that would have been wrong.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The repository carried 22 abandoned agent worktrees under
.claude/worktrees/, left behind by past agent runs. They are
gitignored, so they never reached a commit, but they stayed on disk
and every unqualified repository-wide grep or find walked all 22
copies of the source tree.
Measured before: 48,005 files under .claude/worktrees/, against 856
tracked in the repository. An unqualified search walked 56 times more
files than the repository contains.
The verification gate confirmed both safety conditions before removal:
all 22 worktree-agent-* branch tips were already ancestors of main, and
all 22 worktrees were clean (unmerged: 0, dirty: 0).
Removal steps:
- git worktree remove for each of the 22 worktrees (no rm -rf, so the
registrations in .git/worktrees/ stay consistent)
- git branch -d for each worktree-agent-* branch (lowercase -d, so an
unmerged branch would block deletion instead of being force-deleted)
- git worktree prune to clear administrative entries
Measured after: 0 files under .claude/worktrees/, .claude/ shrank from
4.7 GB to 72 KB. The 856 tracked files are unchanged. HEAD is unchanged
from before the removals. npm run ci exits 0.
This ticket changes no tracked source file. The diff is this ticket
file and the README row, because the work is entirely in gitignored
paths and local branch refs.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
RD-14 asserted that a grep for the four SaveState tags would count at
least 4. All four live on one line of a single-line type declaration, so
the honest answer is 1. The executing agent refused to reformat the type
across four lines to satisfy the number, which would have fought prettier
and violated the ticket's own "move verbatim" decision, and flagged the
check instead.
Use `grep -o | wc -l` when you mean occurrences. Say "lines" when you
mean lines.
This is the seventh miscalibrated acceptance command in this backlog, and
the third an executing agent caught using these rules rather than the
supervisor.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
RD-12 and RD-13 moved both ActionState consumers into their machines. This
left ActionState with zero users, so this ticket deletes it outright.
SaveState is different. It has two four-way consumers (brief.page.ts and
org-template.page.ts) that still switch on all four cases, so it survives.
It moves into debounced-save.ts, next to createDebouncedSave, the only
function that sets it. Both store imports merge into the existing
debounced-save import line. The doc comments that named ActionState are
reworded, since the type no longer exists. No store's public saveState
signature changes, so brief/ui/ needs no edit.
This completes the phase. Two encodings survive: RemoteData for fetched
data, and each machine's own state union. SaveState stays as an explicitly
justified third encoding, for a separate concern (debounced autosave status)
that the other two do not cover.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Before this change, org-template.store.ts held the action lifecycle in an
actionState signal and the publish impact-confirm gate in an independent
pendingPublish signal. The two were representable in combination, so
pendingPublish === true and busy === true could both hold at once. That
state was meaningless: the UI would show the publish-impact confirmation
while a publish was already in flight.
OrgTemplateState.Loaded now carries one action field, a four-variant union
(Idle | ConfirmingPublish | Busy | Failed). ActionStarted overwrites the
field straight to Busy from any prior tag, so ConfirmingPublish and Busy
can never coexist — not by convention, but because one field can only
hold one tag. requestPublish and cancelPublish become dispatches
(PublishRequested/PublishCancelled); as the reducer already no-ops
outside Loaded, this changes no behaviour. The other four commands
(confirmPublish, rollback, proefbrief, flushSave) keep their existing
loaded() guards. busy, lastError and pendingPublish stay on the store as
computed values reading the new union, with byte-identical public
signatures — no file under brief/ui/ changes.
Ran gen:behaviour-spec for the six new reducer cases.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
The sixth acceptance-check miss, and the first the executing agent
caught using these rules rather than the supervisor. RD-12 asserted
`git grep "ActionState" -- apps/ssp/src/app/brief` returns nothing, but
org-template.store.ts lives in that directory and is deliberately out of
scope until RD-13. The check could never pass.
Sharpen the existing habit into a mechanical step: name the files, never
a parent directory.
The agent also caught the RD-11 substring trap in its own work, where the
new type BriefActionState contains ActionState, and word-anchored the
check instead of renaming away from the natural name.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The action lifecycle (Idle | Busy | Failed) lived in an imperative
store-level signal, set from ten call sites outside the reducer. The
reducer could not enforce which action transitions are legal.
Add `action` to `BriefState.Loaded`, driven by three new messages
(ActionStarted, ActionFinished, ActionFailed) and handled in `reduce`.
Replace every `actionState.set(...)` call in `brief.store.ts` with the
matching `dispatch`. `BriefLoaded` resets `action` to Idle, so a fresh
load clears a stale action error instead of letting it outlive the
reload.
`busy` and `lastError` stay as `computed`s on the store with a
byte-identical public signature — they are the render seam for four
components and two page templates, and the union belongs in the
machine, not the components. `revealBigNummer` still sets only
`Failed`, never `Busy` — an existing asymmetry, not changed here.
`SaveState`, `org-template.store.ts`, and `pendingPublish` are out of
scope (RD-13, RD-14).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
RD-11 asserted that `git grep "machineRemoteData\|LoadLifecycle"` returns
nothing, but the replacement it mandates is named `fromLoadLifecycle`,
which contains the old name as a substring. The check could never pass.
When a new name contains the old one, anchor on word boundaries.
Add a fourth habit: prefer a number over a prohibition for anything that
must not change. "Do not rename BriefStatus" invites reasoning around it.
"This count must still be 54, and if it moves, revert rather than adjust
the number" does not. RD-11 renamed tags across 19 files with a wire
contract in the same file, and the count held.
Also record a real trap found during RD-11: stale gitignored
bigregister.db artifacts fail backend tests with "no such table" on an
otherwise clean tree. A 0-byte file months old failed 6 tests.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
`machine-remote-data.ts` defined a third encoding of an in-flight fetch:
`LoadLifecycle`. It had three call sites, all one identical line, and the type
was never imported by name. Move the mapping into `remote-data.ts` as
`fromLoadLifecycle`, beside its neighbour `fromResource` — a `RemoteData`
constructor, not a sixth encoding.
The lowercase `loading`/`failed`/`loaded` tags on `BriefState`,
`OrgTemplateState` and `StamdataEditorState` existed only because
`LoadLifecycle` required them. Now that the constraint is inline and
PascalCase, the three machines' load-lifecycle tags become `Loading`,
`Failed` and `Loaded` — matching their own PascalCase message tags in the
same file. `stamdata-editor.machine.spec.ts` no longer asserts a PascalCase
message producing a lowercase state.
`BriefStatus` (the letter's draft/submitted/approved/rejected/sent status,
parsed off the wire from `BriefViewDto`) is a separate tag family and is
untouched — its tag count stays 54 before and after this change.
Delete `machine-remote-data.ts` and merge its spec into `remote-data.spec.ts`.
Regenerate `behaviour-spec.mdx` (the `machineRemoteData` section heading
becomes `fromLoadLifecycle`) and confirm `gen:snippets` reports no drift, since
`remote-data.ts` carries a showcase region.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
The wizard shell took two inputs to say one thing: a flat WizardStatus
string and a separate errorMessage input. Each wizard needed three
computeds (failedError, errorMessage, shellStatus) to take the state
apart and put it back together for the shell.
WizardPhase replaces both inputs with one discriminated union. Its
Failed variant carries the message directly, so no data travels through
a second channel. Each wizard now maps its own tags onto WizardPhase in
one computed, composing the localized failure prefix at the same spot
errorMessage did before. The three machines and their own vocabulary
(Editing/Answering/Invullen, Indienen/Ingediend/Mislukt) are unchanged;
only the shell's input contract changes.
The shell reads the Failed message via the existing whenTag helper,
because @switch cannot narrow a union in an Angular template.
Both $localize ids (wizard.indienenMislukt, regWizard.indienenMislukt)
keep byte-identical source text, so no locale file changes.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Four acceptance commands in this backlog could not pass as written. The
fourth, in RD-09, reached this backlog's own ticket files and 22
gitignored worktrees, so satisfying it literally would have rewritten the
history of completed tickets.
The general fix is `git grep` instead of `grep -r`: it searches tracked
files only, so untracked and gitignored paths cannot pollute the result.
Measured here, `grep -r` finds 132 hits under .claude/ where `git grep`
finds none. RD-17, RD-18 and RD-19 are repo-wide sweeps and depend on
this.
Also correct RD-09's Order row. It claimed the ticket covered a generator
and a skill file; neither teaches the deleted idiom, as recorded in
PLAN.md.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
RD-05 through RD-08 replaced the hand-called submit method with
createStore's effect map. Two teaching documents still showed the old
method in a code block, as the answer to "how does a submit happen?".
Both blocks also called a function that no longer exists.
Rewrite the code block in ARCHITECTURE.md section 2d and its
fp-tea-atomic-design.md counterpart. Both now show the effect map, keyed
on the Submitting tag, using the same herregistratie worked example with
its optimistic begin/confirm/rollback calls. Both use draftSync.submit,
the call the two herregistratie wizards make today.
State the two properties the old idiom lacked, since they are the reason
for the change: entering a state runs its effect, so a dispatch cannot
skip it; and double-submit protection is structural, because the effect
fires only on a tag transition. Add one sentence on the Seed exemption: a
mount or restore message must not trigger a submit.
Fix the one runIfSubmitting() hop in the write walkthrough at
ARCHITECTURE.md's line 574. The rest of section 6a stays stale on
purpose — RD-31 owns it, including its line citations and dead paths.
fp-tea-atomic-design.md's broken pre-monorepo paths stay stale too —
RD-32 owns those.
Set RD-09's Status to done and its README row to done in the same
commit.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
The three wizards paired a dispatch with a hand-written effect call
(onPrimary/onRetry + runIfSubmitting/runIfIndienen). A missed call failed
silently. RD-05 added the effect map and RD-07 added the Primary message;
this ticket moves each wizard onto both.
Each wizard now registers its submit effect on createStore, keyed by its
own submitting tag (Submitting for herregistratie and intake, Indienen for
registratie — the type catches a wrong key at compile time). The optimistic
begin/confirm/rollback calls stay inside the effect body, unchanged. The
template dispatches Primary and Retry directly, matching how Back already
worked. onPrimary, onRetry, and runIfSubmitting/runIfIndienen are deleted
from all three components.
herregistratie-wizard drops under the 250-rule-line budget, so its
eslint-disable max-lines header is removed in this same commit (RD-02's
self-cleaning mechanism). intake-wizard and registratie-wizard stay over
budget and keep theirs, both already flagged for RD-22/RD-23.
Three doc comments (in the three machine files, plus one in store.ts) named
the deleted onPrimary()/runIfSubmitting() identifiers in prose. Reworded
them so the "idiom is gone from the repo" grep check is not defeated by its
own explanatory comments.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
PLAN.md stated that plop-templates/form-machine.hbs generates
runIfSubmitting, and that .claude/skills/form-machine/SKILL.md teaches
it, so the next scaffolded form would recreate the bug RD-06 fixed.
Both are false. The plop template is machine-only: 74 lines, no
@Component, no runIfSubmitting. The skill never mentions it.
Two real sites remain, both teaching the deleted idiom in a code block:
ARCHITECTURE.md:314 and :574, and fp-tea-atomic-design.md:342. RD-09 is
therefore a two-document fix, still worth doing, but not the urgent
generator fix this plan claimed.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Each wizard component re-derives the step-boundary decision the reducer
already owns: advance on a middle step, submit on the last step. This
ticket moves that decision into the machine, so RD-08 can replace the
component's guard with one dispatch.
Add a `Primary` message to each Msg union, and export a `primary(s)`
function next to the existing `next`/`submit` pair. `primary` is a
three-line branch that delegates to `next`/`submit` and writes no new
validation. Each machine tests "last step" in its own vocabulary, per
the ticket's Decisions block: `herregistratie` checks `step === 3`,
`intake` checks `currentStep(s) === 'review'`, `registratie` checks
`currentStep(s) === 'controle'`. `Next` and `Submit` stay in every
union and every reducer — `Primary` is purely additive.
Add 3 spec cases per machine (9 total): Primary advances from a
non-final step, Primary submits from the final step, and Primary is a
no-op outside the editing state. Each case also asserts the
equivalence the ticket requires for RD-08's migration:
`reduce(s, Primary)` equals `reduce(s, Next)` at a non-final step, and
equals `reduce(s, Submit)` at the final step.
Regenerate `behaviour-spec.mdx` for the 9 new `it()` titles.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
scripts/ci-local.sh gated its two storybook + axe steps on
`[[ "${1:-}" == "--full" ]]`, so the flag had to arrive as a positional
argument. CLAUDE.md documents `npm run ci --full`, but npm parses that
flag itself and exports `npm_config_full=true` instead of passing it
through. Proven with `npm run env --full`.
The documented command therefore skipped both steps and still printed
"local CI passed" — a gate that reported success without running. Anyone
following CLAUDE.md has been skipping storybook-a11y.
Accept either form. Fixing the script makes every existing document
correct, instead of changing four of them and re-teaching the reader.
Found while verifying RD-06, which adds stories and so depends on those
steps actually running.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Before this fix, a failed submit on the besluit-form or the
change-request-form left the user stuck. Both templates rendered `Failed`
through the same `@else` branch as the editable form. In besluit-form the
fields read from `Editing` only, so they went blank. In change-request-form
the fields still showed the sent value, but `SetField` only applies to
`Editing`, so typing did nothing. In both forms the submit button stayed
enabled, but `Submit` is a no-op outside `Editing`. The only escape was a
page reload.
After this fix, `Failed` gets its own template branch: an error message, a
read-only summary of what was sent (an `app-data-block`, reused from the
existing BRP-address pattern), and a "Opnieuw proberen" button that
dispatches `Retry`. Both machines already handle `Retry` (`Failed ->
Submitting` with the preserved data), so no machine change was needed.
Both components also move to `createStore`'s effect map (RD-05): the
`Submitting` handler replaces the hand-called `runIfSubmitting`, so
`onSubmit` is now a single `dispatch`. `runIfSubmitting`/`runIfIndienen`
now remain only in the three wizards, migrated later by RD-08.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
createStore now takes a third, optional StoreEffects map. Each key is a
Model tag. The store runs that tag's handler after update() returns, and
only when the store enters the tag: the previous tag differs from the new
tag, and the message is not Seed (the mount/restore message in every
machine that has one).
This closes the gap where a component had to call dispatch(msg) and then
a private runIfSubmitting() by hand, or state got silently stuck. No call
site changes here. RD-06 and RD-08 migrate the 5 components that
duplicate that pattern today.
The effect map is a conditional type, not a generic constraint, so a
tagless Model (store.spec.ts's plain number store) still resolves it to
never and needs no third argument. Both tag checks use a typeof/in guard
for the same reason.
Regenerated libs/shared/docs/behaviour-spec.mdx for the 5 new spec titles.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
RD-03 added `overzicht: ['registratie']`, so the comment claiming
`herregistratie: ['registratie']` is "the one sanctioned cross-feature
edge" became false. There are two, and both point at registratie.
State what each edge is for instead of counting them, so the next
context added does not make the comment wrong again.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The dashboard page lived inside registratie, a context with no permitted
outbound edges. Two of its sections do cross-context navigation: action
links to herregistratie/intake/brief/concepts, and an admin links section
that imported ADMIN_LINKS directly from the app shell.
Move the page and these two navigation sections into a new overzicht
context. Declare overzicht: ['registratie'] as the one sanctioned edge, so
the four data sections that stay in registratie/ui/dashboard/ remain
reachable. The route stays at /dashboard: it is a user-visible URL and four
e2e specs assert it, so only the import path and class name change.
beheer-links.section.ts now injects HEADER_ADMIN_LINKS, the same token the
site header uses, instead of importing the app shell's nav config directly.
Used npm run gen:context and corrected its three known-wrong outputs: the
overzicht: [] edge, the scaffolded overzicht.page.ts (the real page is the
moved dashboard.page.ts), and the scaffolded path: 'overzicht' route.
Verified the boundary is enforced, not decorative: a temporary import of a
herregistratie file into overzicht.page.ts makes dep:check fail, as
expected, then reverted.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
The dashboard refactor proved a page can be 42 lines. This rule holds
the rest of the app to that budget, before the split work in RD-20
through RD-26 begins.
Add max-lines at 250 (skipBlankLines, skipComments), scoped to
{apps,libs}/**/*.{page,component,section,step}.ts. The glob includes
section and step, the file kinds the dashboard refactor invented, so
new files from this arc do not escape the guard.
Add linterOptions.reportUnusedDisableDirectives: 'error' repo-wide.
ESLint 9 only warns on an unused disable by default, so a stale
exemption would not fail the build. At 'error', every later file
split must delete its own exemption or the build goes red.
Add a dated /* eslint-disable max-lines */ header to each of the
seven files that exceed the budget today, each with a reason and the
ticket that removes it. letter-canvas keeps its disable after RD-26,
because most of its lines are CSS and the rest is one letter.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Convert the approved readable-codebase plan into the ticket protocol this
repo already uses for long arcs (docs/project/backlog/README.md, 75 work
packages driven to done through it).
Adds docs/project/readable-codebase/ with:
- README.md — the ledger: session protocol, GREEN definition, restart
recovery, the agent loop, and an Order table of 35 tickets with their
dependencies and status.
- PLAN.md — the design record, committed so the repository holds the
complete state. A fresh session needs nothing outside it.
Two properties make the arc restart-safe. Each ticket updates its own
Status line and its README row in the same commit as its code, so the
ledger and git log cannot desync. And the plan lives in the repository,
so recovery is three commands and needs no prior context.
Records one correction found while scaffolding: the 22 abandoned agent
checkouts under .claude/worktrees/ are live registered git worktrees, not
orphaned directories. RD-15 must use `git worktree remove` and verify each
branch tip is an ancestor of main first. Spot-checked four; all are
already merged via 637d500.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Catches drift accumulated since the Application → Aanvraag rename
(Step 1/8, Step 2/8) plus the new tests from this branch.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
BrpAddressDto's fields are generated as optional, so returning it
directly from parseBrpAddress lost the narrowing the runtime check
already did. This broke the build once registratie-lookup.store.ts
assigned the parsed address into a stricter local type. Map to a
proper BrpAddress domain shape at the trust boundary instead.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Each dashboard section (Mijn aanvragen, Wat moet ik regelen, Mijn
registratie, Specialismen, Wat wilt u doen, Beheer) now owns its own
store access, async state, and template. DashboardPage becomes pure
composition.
Extract the repeated RemoteData Success-narrowing pattern into
successOf() and the dashboard sort/split logic into
sortForDashboard/concepten/ingediend, both with tests.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
dashboard-view.dto.ts and brp-address.dto.ts each shadowed a generated
type: DashboardViewDto was declared twice (hand-written with required
fields, generated with everything optional), reconciled only by
structural typing. Both are gone.
dashboard-view.adapter.ts now imports the generated DashboardViewDto/
RegistrationDto/PersonDto/RegistrationStatusDto directly. Its parse
does real work now instead of an identity copy: parseRegistrationStatus
validates each status variant's required fields per-tag (the generated
type flattens the union, so a Geregistreerd row missing
herregistratieDatum previously passed the boundary unnoticed — it no
longer does). HerregistratieDecisions moves from contracts/ to
domain/registration.ts, so no contracts-typed value reaches a page.
brp.adapter.ts drops its own BrpAddressDto shadow the same way.
Hand-written contracts/*.dto.ts count: 4 -> 2 (duo-diplomas.dto.ts and
stamdata.dto.ts remain — both parse fields codegen can't type at all).
Part of the dashboard-readability refactor (see the approved plan).
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The wire said Application, the domain said Aanvraag — one aggregate with
two names at every hop. Rename the backend DTOs and the /applications
route to /aanvragen, regenerate the typed client, and rename the frontend
adapter/store to match.
Renamed: ApplicationSummaryDto/DetailDto, CreateApplicationRequest,
SubmitApplicationRequest/Response → Aanvraag* equivalents;
ApplicationsAdapter/Store → AanvragenAdapter/Store;
applications.adapter.ts/applications.store.ts → aanvragen.*.
Left untouched: the admin Case/Zaak vocabulary (/admin/cases,
AdminCasesStore) — a separate read model, not part of this rename; the
internal BigRegister.Domain.Applications namespace and the Applications
EF table (renaming those needs a new EF migration, out of scope here).
Part of the dashboard-readability refactor (see the approved plan).
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
docker compose up at the repo root does not start OpenZaak. A new reader
could easily assume it does. Add a short section to
backend/openzaak/README.md that shows how to check the four containers and
how to curl OpenZaak directly, using the same probe
bootstrap-catalogus.sh already relies on. Add one line to the root README
pointing there.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Closes the CD refactor backlog (docs/project/refactor-backlog-setup/refactor-backlog/99-backlog.md).
All 33 code tickets and the four gated ADR-fixes (ADR-C-001, ADR-C-003,
ADR-C-007, ADR-C-009) are merged, one commit per ticket, across six CD
batches plus the ADR-fix batch. npm run ci is green after every merge in
the arc, each verified independently.
Highlights: RB-01/02 fixed a BSN leak in the persisted audit trail and an
unauthorized document-content endpoint. RB-09/13 landed Session -> Principal
per ADR-0002. RB-12 added a route-table authz gate as a CI safety net.
RB-19 reordered the backend's 940-line Program.cs into reads-then-writes,
verified as a pure move by comparing every (route, gate, handler) triple
before and after. RB-24..30 moved libs/shared/upload into its proper
layers and made every layer testable. RB-31 found and fixed a real
ADR-0006 violation: two tests asserted a wizard state the real reducer
cannot produce.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
All three tickets RB-31 to RB-33 merged, one commit per ticket. RB-31 found a
genuine ADR-0006 violation: two registratie-wizard tests asserted a state the
real reducer cannot produce. RB-32 closed ADR-0003's own predicted failure
mode with a permanent CI drift guard rather than a one-time fix. RB-33 chose
deletion over adoption for an unused test helper, since manufacturing a first
caller would have removed no real duplication.
This closes the CD implementation phase. All 33 code tickets and the four
gated ADR-fixes are merged; npm run ci is green after every merge in this
arc, each verified independently rather than trusting an agent's own report.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
ADR-C-010: intake, registratie-wizard, besluit and brief machine specs
hand-rolled a state literal, three of them hardcoding errors: {} by hand
instead of running real Msgs through the real reduce (ADR-0006 §2). intake
now shares the existing intake.testing.ts with the acceptance spec instead
of ignoring it; the other three each get a *.testing.ts one-liner.
Found real drift: two registratie-wizard tests asserted a cursor-2 state
reached before any diploma was chosen, which the real reducer cannot
produce (advancing past beroep requires a diploma already set). Replayed at
cursor 1 instead; submit() validates the whole draft regardless of cursor,
so the assertions are unchanged.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
# Conflicts:
# docs/project/refactor-backlog-setup/refactor-backlog/99-backlog.md
Four machine specs built their starting state with a hand-rolled object
literal instead of replaying real Msgs through the real reduce, the
exact anti-pattern ADR-0006 section 2 forbids. Three of the four also
hardcoded errors: {}, a shape the reducer might never actually produce.
intake.machine.spec.ts now imports the existing givenIntake from
intake.testing.ts (previously used only by intake.acceptance.spec.ts).
Three new one-line *.testing.ts files export the same given(reduce,
initial) wrapper for registratie-wizard, besluit, and brief. Every old
literal helper (answering, invullen, editingWith, loaded) is replaced
by a message replay that reaches the same state.
Two tests in registratie-wizard.machine.spec.ts asserted a cursor value
the real reducer cannot reach (cursor 2 with no diploma chosen yet,
which requires a diploma to already be set). Both are re-pointed at the
reachable cursor-1 equivalent; submit() validates the whole draft
regardless of cursor, so no assertion changed. Recorded in
implementation/rb-31.md, not worked around.
No *.machine.ts reducer was touched. All four specs pass; npm run ci
is green (lint, typecheck, dep:check, format, tokens, seam, all four
test suites, both app builds, backend 293/293, api-client drift).
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
ADR-C-008: 9 files carried a CIBG-GAP EXTENSION marker against 8 register
rows. language-switcher had a well-formed marker and no row. Adds the row,
plus an optional ~14-line drift guard in check-tokens.sh that diffs the
marker set against the register and fails naming any missing row, so this
drift cannot silently reoccur.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
# Conflicts:
# docs/project/refactor-backlog-setup/refactor-backlog/99-backlog.md
The register at libs/shared/docs/cibg-gaps.mdx had 8 rows for 9
CIBG-GAP EXTENSION markers in code. language-switcher carries a
well-formed marker with no matching row, exactly as ADR-C-008 and
adr-c-007.md's handoff note flag. Add the row from the component's
own marker comment.
Also add a small guard to check-tokens.sh (folded into check:tokens,
as ADR-C-008 suggests as an optional step): it diffs the marker set
in code against the register's rows and fails CI on drift. Verified
working with a scratch marker file before removing it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
ADR-C-011: unwrapOk had zero consumers anywhere in apps/ or libs/ since it
shipped, and the one candidate call site (submit-change-request.spec.ts's
inline parse-and-throw guard) already satisfies ADR-0006 section 3's real
requirement, never a cast. Manufacturing a first caller to justify keeping
the helper would remove no real duplication. Deleted rather than adopted.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
unwrapOk had zero consumers in apps/ or libs/ since ADR-0006 shipped it.
The one call site the finding named already satisfies the ADR's real
rule (call the real parser, never a cast) with an inline guard, so
adding a manufactured first caller was not the better fix. This commit
deletes the helper and its file, and updates the one doc sentence that
named it. The finding's call site is unchanged. See rb-33.md for the
full adopt-or-delete reasoning.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
All seven tickets RB-24 to RB-30 merged, one commit per ticket. Records the
actual wave split, since the backlog's own depends-on column missed that
RB-24 rewrites imports in two of RB-28's target files.
RB-24 expanded its own scope to fix a second, real boundary violation that
deleting its acceptance criterion exposed, reviewed and accepted. Two more
findings were shown stale or overstated, on top of the nine from earlier
batches. RB-26 and RB-27 both correctly declined part of their own ticket's
proposed shape.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
TE-005: xhrUpload buried the 2xx-vs-not check, JSON.parse-with-fallback and
ProblemDetails mapping inside XHR listener bodies, unreachable without
stubbing the XHR global. uploadOutcome(status, responseText) is now a pure
function with no DOM and no XHR stub in its spec. Abort-vs-error
disambiguation stays where it is: it fires on a different event with no
status or responseText, so it cannot fit the extracted signature. The
optional currentScenario() move into KeepaliveTransport.send() was not
taken, since it would cross into upload-shell.service.ts, outside this
ticket's scope.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
UploadAdapter.xhrUpload built new XMLHttpRequest() directly and put the
actual decisions inside its load listener: 2xx-vs-not, JSON.parse of the
body with a fallback, and ProblemDetails mapping via parseError. None of
it was reachable without stubbing the XHR global, so it had no spec
(TE-005; file LH 5/64, BRH 3/57).
Extract uploadOutcome(status, responseText): Result<string, {
documentId }>, a pure function next to genericError/parseError. It holds
the 2xx check, the JSON.parse-with-fallback, and the ProblemDetails
mapping. The load listener is now a two-line dispatch into it.
Abort-vs-error disambiguation stays where it is: it decides whether a
response exists at all, before uploadOutcome would even run, and the
proposed signature has no field for "aborted". It is already a one-line
ternary with no DOM-only logic to extract.
Add upload.adapter.spec.ts: plain describe/it, no DOM, no XHR stub,
covering a 2xx success, a 2xx unparseable body, a non-2xx ProblemDetails
body, a non-2xx non-ProblemDetails body, and the 200/300 boundary.
Verified red by editing uploadOutcome down to one line (an Edit, not
git checkout): 4 of 5 new specs failed. Re-applied with a second Edit.
Coverage for upload.adapter.ts: LH 5/64 -> 12/65, BRH 3/57 -> 7/59.
Skip TE-005's optional half (moving the currentScenario() branch into
KeepaliveTransport.send()): it needs a second file, upload-shell.
service.ts, and this ticket's own scope fences it to upload.adapter.ts
and its spec. The dev simulator's behaviour is unchanged.
Mark RB-27 implemented in 99-backlog.md and add its implementation note,
including a batch 5 close-out.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
TE-006: StamdataStore.download(), BriefStore.previewLetter() and
OrgTemplateStore.proefbrief() each ended in raw DOM blob calls jsdom cannot
meaningfully execute, so their success paths were unassertable and
download()'s two-clause guard true-branch was permanently dark.
BLOB_PRESENTER mirrors the SESSION_PORT shape; all three commands go through
it. download()'s branch coverage goes from 40.5% to 67.6%, and
org-template.store.ts gets its first spec at all.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
# Conflicts:
# docs/project/refactor-backlog-setup/refactor-backlog/99-backlog.md
# libs/shared/docs/behaviour-spec.mdx
Three application-layer commands ended in raw DOM calls (URL.createObjectURL,
window.open, document.createElement('a').click(), URL.revokeObjectURL) as
their last statement. jsdom cannot assert a call that is also the end of the
function, so each command's success path stayed unassertable, and
StamdataStore.download()'s two-clause guard stayed permanently dark on its
true branch (TE-006).
Add BLOB_PRESENTER (libs/shared/src/application/blob-presenter.ts), an
InjectionToken mirroring SESSION_PORT's shape: an interface with open()/
download(), a real implementation preserving the existing open()-never-
revokes vs download()-always-revokes asymmetry, provided in root. Route
StamdataStore.download(), BriefStore.previewLetter(), and
OrgTemplateStore.proefbrief() through it.
Add specs with a recording fake presenter: StamdataStore.download()'s guard
(both clauses) and its success path, asserting toJson(...)'s exact output
reaches the file; BriefStore.previewLetter()'s existing success test now
goes through the seam instead of spying on window/URL directly; a new
org-template.store.spec.ts (none existed before) covers proefbrief()'s
success and failure paths.
Verified red without the fix by editing the download() filename to the
wrong extension, watching the success-path spec fail, then restoring it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
TE-003: UploadShellService documented UploadTransport as the swap seam, then
bound the concrete, unexported KeepaliveTransport class directly, so a spec
could not fake it. UPLOAD_TRANSPORT copies the SESSION_PORT shape; the
default factory returns the same instance, so runtime behaviour is
unchanged. upload-shell.service.ts goes from 0% to 88.57% line coverage
across 16 new specs for upload(), cancel(), delete() and pollReturning().
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
# Conflicts:
# docs/project/refactor-backlog-setup/refactor-backlog/99-backlog.md
# libs/shared/docs/behaviour-spec.mdx
UploadShellService injected the concrete KeepaliveTransport class instead
of a token. The class was not exported, so a spec could not fake it, and
could not provide against the UploadTransport interface either, since an
interface is not a DI token. The port existed only on paper.
Add UPLOAD_TRANSPORT, an InjectionToken with a default factory that
resolves the same KeepaliveTransport singleton, following the
SessionPort/SESSION_PORT shape. UploadShellService now injects the token.
Runtime behaviour is unchanged.
Add upload-shell.service.spec.ts: a recording fake transport plus a fake
UploadAdapter exercise upload(), delete(), cancel() and pollReturning(),
the four methods the missing seam left unreachable. Coverage for
upload-shell.service.ts goes from 0% to 88.6% line / 85% branch.
Mark RB-25 done in 99-backlog.md and add its implementation note.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
TE-004: createUploadController performed three inject() calls, an effect()
registration and a window listener before returning, so the real policy
buried inside it — deciding per file whether to reject or start an upload —
was reachable only through a TestBed. planFileSelection in upload.machine.ts
is now that decision as a pure function taking plain {name, type, size}
objects; the controller executes the plan and keeps the one impure step
(crypto.randomUUID()) it can't move. No change to the controller's public
surface.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
# Conflicts:
# docs/project/refactor-backlog-setup/refactor-backlog/99-backlog.md
# libs/shared/docs/behaviour-spec.mdx
createUploadController required inject(), an effect(), and a window listener
before a test could reach it. The file-selection policy trapped behind that
cost now lives in a pure function, planFileSelection, in upload.machine.ts.
planFileSelection takes plain { name, type, size } objects, not File, and
decides per file whether to reject it or accept it, with no I/O. The
controller executes the plan: it dispatches a rejection as-is, and starts the
upload for an accepted file (the one step that needs crypto.randomUUID()).
A new spec covers the three outcomes: the 'multiple' batch rejection, a
rejectReason-based rejection, and the accept case, plus order in a mixed
batch. Verified red-then-green with a temporary stub, undone by a second edit.
No change to the controller's public surface or to the calling organism.
previewUrlFor (added by RB-24) is untouched.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
TE-008: five guard decisions in BriefStore (Save, Submit, Send, the shared
Approve/Reject review path) were pure functions of status tag, actor role and
entity completeness, but each sat inside a lock-held, DB-opening method, so a
spec could not exercise a decision without a booted host and a real SQLite
file. BriefRules.cs holds the five pure statics; BriefStore keeps its lock,
its Db.Create(), its static shape and every method signature. 29 new
free-running unit assertions in BriefRuleTests.cs; the existing host-booting
brief endpoint tests are unchanged.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
# Conflicts:
# docs/project/refactor-backlog-setup/refactor-backlog/99-backlog.md
# libs/shared/docs/behaviour-spec.mdx
TE-007: Render already accepts the letter's instant and uses it correctly for
the letterhead, but the body's datum placeholder resolved through ResolveAuto,
which ignored at and read DateTimeOffset.UtcNow. Threaded at through
RenderParagraphs and RenderNode, both already in Render's call chain with at
in scope. Zero public API change, zero call-site change. The bug this
prevents: re-rendering an archive or back-dating a letter would otherwise make
the letterhead and body dates disagree within a single document.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
# Conflicts:
# docs/project/refactor-backlog-setup/refactor-backlog/99-backlog.md
LetterHtml.Render already receives the letter's instant and uses it
for the letterhead date. The body's "datum" placeholder resolved
through ResolveAuto, which ignored that instant and read the wall
clock instead. This is not a shipped bug today, because every current
caller passes Now() at render time. It becomes one the moment Render
runs with a historical instant (an archive re-render, a back-dated
letter): the letterhead and the body would then disagree within one
document.
Thread the existing "at" parameter down through RenderParagraphs and
RenderNode into ResolveAuto's "datum" case. Render's own signature,
and every call site, stays unchanged.
Add two tests with a fixed historical "at": one pins the body's
rendered date to the expected Dutch string, the other asserts the
letterhead date and the body date agree. Both fail red against the
old code, showing today's date instead of the pinned one.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
BriefStore's five guard decisions (Save, Submit, Send, and the shared
Approve/Reject review path) were pure functions of status tag, actor role,
and entity completeness, but each sat inside a lock-held, DB-opening
method. A spec could not exercise the decision without a booted host and
a real SQLite file.
Extract the guards into a pure Domain/Letters/BriefRules.cs. BriefStore
keeps its lock, its Db.Create(), its static shape, and every method
signature — only the if cascades move. Add BriefRuleTests.cs (29
assertions, ~120 ms, no host boot) covering every branch, including the
rejected-to-draft reopen on save, the required-filled gate on submit,
and the non-drafter and self-review denials. The existing host-booting
brief endpoint tests are unchanged and still pass, proving the
extraction preserved behaviour.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
ADR-C-002: libs/shared/src/upload/ held a network adapter outside
infrastructure/ and the only Elm machine outside a domain/ folder, and the
dependency-cruiser rule was written around the violation rather than the
violation being fixed. The five files move to infrastructure/, domain/ and
application/, and the ^libs/shared/src/upload/ carve-out is gone.
Deleting the carve-out exposed a second, real violation that the old path had
hidden from the ui-not-infrastructure rule: three UI components injected
UploadAdapter for nothing but a one-line wrapper over its own exported pure
uploadContentUrl. They now read previewUrlFor from their application-layer
collaborator. dep:check passes for both apps with the clause removed, which is
the ticket's acceptance criterion.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
libs/shared/src/upload/ held a network adapter, an Elm machine, and two
application-layer coordinators outside the folder-per-layer convention every
other context follows. The dependency-cruiser rule carved an exception around
the misplaced adapter instead of the violation being fixed.
Move all five files to the layer each belongs to (git mv), update every
import across 24 consumer files, then delete the carve-out clause from
.dependency-cruiser.base.js. No export renamed, no file split, no spec
content changed.
Deleting the carve-out exposed a second, pre-existing rule violation:
ui-not-infrastructure had never fired against upload.adapter.ts because its
old path did not match /infrastructure/. Three UI components injected
UploadAdapter directly for its one-line contentUrl() wrapper. Route each
through the existing pure uploadContentUrl() function via the application
layer (upload-controller's new previewUrlFor, OrgTemplateStore's new
previewUrlFor) instead — the same idiom brief.store.ts already used.
npm run ci passes; dep:check is clean for both apps with the carve-out gone.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
All six tickets RB-18 to RB-23 merged, one commit per ticket. Records the two
incomplete tickets that the agents reported, RB-22's deliberate departure from
the runResult idiom, and how RB-19 was verified as a pure reorder.
Adds five dispatch lessons. The stale worktree base is now the rule at 11 of 13
agent-runs. A spend limit killed four agents mid-flight and a message resumed
each one from its own transcript, so no work was redone.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
CQ-006: the file declared direction as its organising principle, then switched
to feature grouping without saying so, and five sections interleaved reads and
writes. Each section now orders reads first, with the WP-65 sub-banner pair.
DELETE /admin/cases/{id} and GET /admin/audit move up beside GET /admin/cases,
129 lines closer. The org-template preview moves to the org-template section.
Pure reordering. Verified centrally: the sorted list of all 47 route strings is
identical before and after, and so is every (route, .Gate marker, wrapper called
in the handler) triple. The swagger.json and api-client.ts diffs are ordering
only.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
CQ-006 found that Program.cs states a reads-then-writes principle at the
top of the file, then abandons it for five feature sections that mix GET
and mutating endpoints in mapping order. This is a pure reorder: within
Document upload, Applications, Admin cases, Brief, and Organization
templates, every GET now precedes every POST/PUT/DELETE, each split by a
`--- reads ---`/`--- writes ---` sub-banner in the style WP-65 already
established for Beoordeling/Besluit.
DELETE /admin/cases/{id} and GET /admin/audit move up beside GET
/admin/cases, closing the 129-line gap CQ-006 measured. GET
/admin/org-template/{subOrgId}/preview moves from the Brief section to
the Organization-templates section it actually belongs to.
No route, signature, DTO, or handler body changed. Every block was cut
by exact line-range slicing, never retyped. The sorted list of mapped
HTTP-method-plus-path strings is byte-identical before and after; every
.Gate(...) count is unchanged; the three routes that moved with a gate
were checked by eye against the wrapper their handler actually calls,
per RB-12's stated limitation that the route-table test only proves a
marker is present, not that it still matches the handler.
npm run gen:api regenerated backend/swagger.json and
libs/shared/src/infrastructure/api-client.ts; both diffs are ordering
only (sorted-file diff is empty), committed alongside per the ticket's
own guidance.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
CQ-007 contract half. GET /brief was the only backend endpoint where a GET
performed a persisted write, and the FE retries GETs automatically, so a
transient failure could enter the create path. BriefStore.GetOrCreate splits
into Get plus the existing ResetAndCreate. GET /brief/preview shared the same
call site and gets the same treatment. RB-22 already made the FE tolerate the
404, so the pair is complete.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
GET /brief allocated a row on first call (BriefStore.GetOrCreate) — the
one endpoint in the backend where a read performed a persisted write.
The FE retries GETs automatically, so a transient failure could enter
the create path more than once; a lock prevented a duplicate row, but
the safety depended on the lock, not on the endpoint being a query.
Split GetOrCreate into Get (a pure query) and the already-existing
ResetAndCreate (POST /brief/reset owns creation). GET /brief now 404s
when the owner has no brief yet. GET /brief/preview used GetOrCreate
too, so it gets the same Get + 404 treatment, forced by the split.
RB-22 already made BriefStore.load() on the FE tolerate a 404 by
calling reset() once; this ticket is what makes that branch live.
Updated the brief/preview/org-template backend tests that assumed
GET seeded a brief on first call to create one explicitly first, and
added a test that GET 404s and writes no row without the fix (verified
red beforehand). Regenerated the API client (npm run gen:api).
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
CQ-007 expand half. BriefStore.load() treats a 404 as 'no brief yet' and calls
the existing reset() command once. load()'s error channel becomes the
BriefLoadFailure union, because runResult folds the HTTP status away and the
store needs it. Today's backend never 404s, so the branch is a no-op until
RB-23 lands the contract half.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
# Conflicts:
# docs/project/refactor-backlog-setup/refactor-backlog/99-backlog.md
# libs/shared/docs/behaviour-spec.mdx
BriefStore.load() now treats a 404 from GET /brief as "no brief exists
yet" and calls the existing reset() command once, instead of showing
the generic load-failed error. BriefAdapter.load() gains a
BriefLoadFailure error channel (notFound | error) so the store can
tell a 404 apart from every other failure; every other adapter method
stays on runSubmit, unchanged.
The once-only bound is a field on the store, not a comment: a second
404 (from a later load() call) always falls through to the ordinary
error path, and the recovery path never calls load() again, so no
loop can form.
This is the expand half of CQ-007's split (04-cqrs-light.md). Today's
backend never 404s GET /brief, so the new branch is dead code until
RB-23 (the backend contract half) ships in a later merge.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
CQ-002: ApplicationsStore.cancel and AdminCasesStore.delete reached the raw
ApiClient and swallowed the failure in a bare catch, so a failed cancel made the
row reappear with no message. Both now fold through runSubmit and expose
lastError, which the two pages render.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
# Conflicts:
# docs/project/refactor-backlog-setup/refactor-backlog/99-backlog.md
# libs/shared/docs/behaviour-spec.mdx
BIO-018: the store was a process-global dictionary keyed on the client-supplied
Idempotency-Key alone, so one caller could replay another caller's key and
receive their cached response. The key is now scoped with the caller SubjectId.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
# Conflicts:
# docs/project/refactor-backlog-setup/refactor-backlog/99-backlog.md
# libs/shared/docs/behaviour-spec.mdx
ApplicationsStore.cancel and AdminCasesStore.delete rolled an optimistic
write back on failure but showed no message — a bare catch with no
Result and no error channel (CQ-002). Both now call runSubmit and set a
lastError signal on failure, mirroring createSubmitChangeRequest in the
same folder. Each page renders the error with the existing app-alert
atom, the same pattern brief.page.ts already uses for lastError.
Added a spec file for ApplicationsStore (none existed) and extended
AdminCasesStore's spec, each asserting the rollback AND the surfaced
error. Verified both new assertions fail without the fix (an Edit
undo/redo of the store method, not git checkout).
Regenerated libs/shared/docs/behaviour-spec.mdx (gen:behaviour-spec) to
pick up the new/renamed test names. Marked RB-20 done in 99-backlog.md
and recorded the change in implementation/rb-20.md.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
CQ-001: createDraftSync was the longest function in the repo and owned three
query paths next to its write path. findConcept and loadConcept are now free
functions that take the adapter, so they have a direct spec without TestBed.
The closure state (id, ensuring, resumeGate) stays where it was, because the
coupling is load-bearing.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The architect approved the four ADR-fix tickets. All four change what the
architecture documents claim. No code changes.
ADR-0001, ADR-C-001: the worked example claimed the POC has no real backend.
It rewrites against `backend/src/BigRegister.Api`. Every path it named is
repointed. The out-of-scope list drops two discharged bullets: 33 `parse*`
boundaries exist, and `npm run gen:api` is real.
ADR-0001, ADR-C-003: a new section states that the generated client is the wire
contract. A hand-written `contracts/*.dto.ts` is the exception for two cases
only. The four survivors stay, because NSwag emits every property as optional
and flattens `RegistrationStatusDto` into five optional strings. The `parse*`
trust boundary stays mandatory, because a generated type is a compile-time
claim about the wire and not a runtime guarantee.
ADR-0003, ADR-C-007: four paths moved in WP-67 and are repointed. Point 4 kept
the principle and changed its example to `skeleton` and `spinner`. Two of its
claims were false and the amendment says so: `app-alert` wraps the vendored
`.feedback` classes, and `site-header` composes the vendored `.titlebar`.
ADR-0004, ADR-C-009: the exception section states a four-part test instead of
one named exception. `OrgTemplateStore` and `FeatureFlagStore` both pass it. RB-07
gated this ticket, because clause 4 needs an audited allow path. RB-07 landed
that, so the ADR does not ratify a control that the code lacks.
Three tickets need a matching CLAUDE.md correction in the same diff. CLAUDE.md
section 2 loses the false `alert` example. Section 4 gets the generated-client
rule and the four-part test.
Two findings were wrong. ADR-C-001 asked to keep an out-of-scope bullet that
reads "SessionStore is in-memory". The session persists to `localStorage` now,
so the bullet covers multi-tab sync only. ADR-C-007 flagged one half of point 4
and missed that the other half is equally false.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
IdempotencyStore keyed a replayed submission on the raw Idempotency-Key
header alone. Two different callers who send the same header value
shared one cache slot: the second caller received the first caller's
cached reference instead of running its own submission.
Program.cs now composes the key as "{SubjectId}:{idemKey}" in the
Submit helper, so the cache is scoped per caller. Add a test that
proves a caller cannot replay another caller's idempotency key and
receive their cached result.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
createDraftSync mixed a read path (findConcept, load, the read half of
resume) with its write path (ensureId, flush, submit, reset) in one
187-line function -- CQ-001's finding. Move findConcept and loadConcept
into a new application/find-concept.ts as free functions that take the
adapter, so they get a direct spec with no Angular TestBed.
createDraftSync keeps the closure state (id, ensuring, resumeGate) and
the whole write path unchanged -- this is a move, not a redesign. The
resumeGate coupling that lets the write path wait for the read path
stays exactly where it was.
createDraftSync shrinks from 187 to 169 lines. draft-sync.spec.ts is
unchanged -- it never called resume()/load() directly, and its 409
recovery test for submit() still exercises the extracted findConcept
through ensureId's catch branch.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Adds a Conventions rule for Simplified Technical English (ASD-STE100). It
covers documentation, code comments, commit messages, ADRs, and the backlog
notes. STE is a controlled language. It makes text easy to read for people
who do not have English as a first language, and easy to translate. The
readers of this project are mostly non-native English readers.
The rule states that STE governs form, not content. Split a long sentence.
Never remove a caveat, a measurement, or a precise term to make text shorter.
The rule does not apply to Dutch identifiers, $localize copy, quoted output,
or existing documents that you are not already editing. It therefore does not
change the Naming convention above it, which keeps domain contexts in Dutch.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
All six merged, gate green (14 steps, backend 260/260). Records the three
tickets that could not be built as written — RB-12's wrapper/public binary
does not fit the route table, RB-14's command exits 0 on a High advisory, and
RB-15 needed a third environment name because RB-09 makes Production fail to
boot — plus RB-13's measured duplication drop (168 -> 32 lines per side).
Adds a section on dispatching implementation agents. Four of six agent-runs
were handed a worktree branched from a stale ancestor; batch 3 was three for
three. That, the background-task parking, and the git-checkout-destroys-work
trap are all cheap to prevent in the prompt and expensive to discover.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
ADR-0002 SS3 models Zorgverlener/Medewerker as different Principal
variants with different login flows. Actor #2 (apps/behandelportal)
landed in WP-61/67 and the union never followed: grep -rn "Principal"
returned one hit, a comment. Both apps' auth/domain/session.ts stayed
byte-identical (`{ bsn, naam }`), so the backoffice's Behandelaar
carried a BSN and logged into the backoffice as a citizen, by DigiD,
under a fabricated citizen's name (login.page.ts). The divergence
ADR-0002 predicted took an orthogonal side door instead
(medewerker.interceptor.ts's X-Medewerker/X-Rollen stamp, which never
touches SessionStore) -- which is why ssp/auth and bhp/auth still
measured as 100%/84% duplicated after ADR-C-006 shared the route
guards. RB-09 (landed the day before) made the backend's
IIdentityProvider able to say "no identity" and fail closed; this
ticket is its named FE half.
Each app's auth/domain/session.ts becomes principal.ts, holding the
one Principal variant that app actually has an actor for: ssp keeps
`{ kind: 'zorgverlener', bsn, naam }` (G1 still strips the BSN before
persisting); behandelportal gets `{ kind: 'medewerker', medewerkerId,
naam, rollen }` (no BSN to strip -- G2 shape validation only). A new
MedewerkerAdapter replaces DigidAdapter in behandelportal, resolving
the existing MEDEWERKER_ID/currentRollen() dev stand-in into a
Principal; because there is no credential to check, it returns
Principal directly rather than a Result whose error variant could
never occur. login.page.ts stops being a BSN/wachtwoord form -- one
explainer line and an "Inloggen met SSO" button -- and its dead
error-handling branch goes with the Result wrapper that justified it.
Measured with tools/baseline-scan.mjs --dup: auth duplication drops
from 168/168 (ssp) and 168/200 (bhp) to 32/179 and 32/259 -- under the
backlog's <40 target. What remains is the ADR-C-006 route-guard
re-export (deliberately identical), generic test/story-file
boilerplate, and one shared fragment of the root-singleton-store
idiom -- not re-converged identity or login-flow logic. SS3's
prediction that the two actors would authenticate differently enough
to justify not sharing auth has now actually been tested, not just
asserted, and held.
Also: renamed Session.bsn to Principal.bsn in two doc comments
(libs/shared/src/infrastructure/subject.ts, subject.interceptor.ts)
that cited the old type name; regenerated
libs/shared/docs/behaviour-spec.mdx (generated file, per its own
banner); recorded the resolution in ADR-0002 as a new amendment,
replacing its "Known debt" section.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
New backend test classes (RouteInventoryTests, SwaggerGateTests) plus
one added case to StamdataEndpointTests.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
BIO-019: GET /stamdata/{table}?peildatum= called DateOnly.Parse
directly, which throws FormatException on anything unparseable — an
unhandled 500 (leaking exception detail in Development) instead of
the 400-with-problem-details every other bad-input check in this file
returns. §3c named backend/Stamdata's 71.7% branch coverage (BL-005)
as the weak spot this bug lived in.
Switched to DateOnly.TryParse; an unparseable value now returns
Results.Problem(detail: ..., statusCode: 400), matching the shape the
upload/change-request endpoints already use. Endpoint doc gained
.ProducesProblem(400), so the OpenAPI doc + generated client were
regenerated and committed in this same diff (RB-09's note records a
prior incident where a response-shape change shipped without this and
the drift went unnoticed).
No FE change needed: libs/beheer's stamdata adapter already funnels
every call through runSubmit, which folds any thrown ApiException
(now including this 400) into a generic Result error — ADR-0001's
"the FE renders the decision" already covers "the server rejected
this input".
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
BIO-015: app.UseSwagger()/app.UseSwaggerUI() ran unconditionally, so
the full OpenAPI document (every route + request/response shape) and
SwaggerUI's interactive "Try it out" were reachable in every
environment, including a real deployment.
Both now run only inside `if (app.Environment.IsDevelopment())`.
AddSwaggerGen/AddEndpointsApiExplorer stay unconditional — DI
registration only, no HTTP surface by itself.
RB-09 already made a non-Development environment throw at startup,
which broke `npm run gen:api` until that script pinned
ASPNETCORE_ENVIRONMENT=Development for its one CLI invocation. This
change sits in the same pipeline, so it was verified rather than
assumed: `dotnet swagger tofile` resolves ISwaggerProvider straight
out of DI and never sends an HTTP request through this middleware, so
gating it can't affect that tool by construction. Ran the real
`npm run gen:api` to confirm — exit 0, regenerated files byte-identical
to what's committed.
New tests exercise the gate on a third ("Staging") environment name,
not Production — Production already can't boot at all post-RB-09, so
a Production-environment test would only re-prove that unrelated
startup throw, not this gate.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
BL-006: the backend has zero automated architecture enforcement.
BIO-016 names the concrete consequence for authorization — nothing
asserted the *set* of gated endpoints, so BIO-003's X-Admin gate
(outside Authz) and BIO-004's two ungated endpoints were caught only
by a human reading Program.cs, not by CI.
Adds RouteInventoryTests: walks the real app's EndpointDataSource and
asserts every mapped route either carries a .Gate("XAdmin") metadata
marker (added at the 16 call sites that already call one of the five
admin wrappers — OrgAdmin/StamdataAdmin/CasesAdmin/Beoordelen/
FlagsAdmin) or appears in a written-down, reasoned allow-list. Proved
it's hard to fool by adding a throwaway unguarded route, watching the
test go red, and reverting.
The allow-list is not "public routes" as the ticket's shorthand put
it — 19 of its 31 entries are ownership-scoped inline (ctx.Zorgverlener()/
ctx.Caller()) endpoints, not public ones, and labelling them public
would misrepresent the exact property BIO-004 was about. Each entry
instead carries its own reason. Implementation note has the full
route-by-route breakdown and judgement calls.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
runSubmit did two things at once: fold a call into a Result, and mint
an Idempotency-Key for it. Five call sites are reads and had no
business minting one — brief.adapter.ts:load, org-template.adapter.ts
:list/:load, and stamdata.adapter.ts:list/:load. stamdata.adapter.ts's
own docstring already said "Both endpoints are reads … There is no
write method" while both called runSubmit; that mismatch is the
sharpest evidence, and the reason the baseline's original "~13
mutations" count (derived from the helper's name, not the code) was
wrong by five in one direction.
Split submit.ts in place: runResult is the try/catch + problemDetail
fold with no mint; runSubmit is runResult wrapping
withIdempotencyKey. Zero behaviour change for the 8 real mutations
(brief save/submit/approve/reject/send/reset, org-template
save/publish/rollback) — same fold, same mint, same timing. The five
reads now run the fold with no pendingIdempotencyKey touched.
submit.spec.ts asserts the split behaviourally via
currentIdempotencyKey() (two reads inside the same call agree only
when a key was minted and reused) rather than mocking a relative
import, matching this repo's existing vitest convention. Verified red
without the fix by temporarily reintroducing the mint into runResult.
ApplicationsStore.cancel/AdminCasesStore.delete (RB-20) and
FeatureFlagStore.set are out of scope and untouched — the latter
already calls runSubmit correctly.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
npm audit --omit=dev gates the shipped frontend bundle; nothing equivalent
existed for the backend, so the entire .NET dependency tree — direct and
transitive — was unscanned (BIO-016 lists it first under "Absent").
The ticket's literal wording would not have worked. `dotnet list package
--vulnerable` is a reporting command: it prints the advisory table and exits
0 regardless. Verified with a throwaway project on System.Net.Http 4.3.0 —
severity High, GHSA-7jgj-8wvc-jh57, exit code 0. A bare `- run: dotnet list
package --vulnerable` would have added a line that reads like coverage in a
compliance review and enforces nothing, which is worse than leaving the gap
visible.
scripts/dotnet-audit.sh runs the scan and matches "has the following
vulnerable packages" — the exact sentence dotnet prints per project on a hit.
One script, two callers (ci.yml and ci-local.sh), so the workflow and the
local gate cannot drift apart.
No severity threshold and no suppression list: picking either before a real
advisory forces the question would be guessing at a policy nobody needs yet.
Secret scanning, BIO-016's other named absence, stays on the checklist.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
All five tickets merged and green on the fixed gate (13/13 steps, exit 0).
RB-07 unblocks ADR-C-009; RB-09 unblocks RB-13 in batch 3.
Adds a "Gate integrity" section recording that every earlier "ci green" in
this file predates the ci-local.sh errexit fix and is weaker than it reads.
Batch 1 has not been re-verified under the honest gate, and the note says so
rather than leaving a reader to assume it was.
Also records what batch 2 leaves open: RB-01's residual is NOT solved by
RB-09 (the upload-content link is still a plain browser navigation with no
credential), and a non-Development non-Production environment fails fast at
GetRequiredService rather than at RB-09's deliberate throw.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Under `set -e` bash exempts every command of an AND-OR list except the last,
so `npm run gen:api && git diff --exit-code ...` silently swallowed a CRASH
in gen:api: the diff never ran and the script sailed on to print "local CI
passed". Verified directly — `bash -c 'set -e; false && true; echo hi'`
prints hi and exits 0, while `false; true` exits 1.
This was not hypothetical. It hid a real gen:api crash introduced by RB-09
(dotnet swagger's design-time host defaults to Production, which that ticket
made throw at startup). .github/workflows/ci.yml would have caught it, since
it runs each step as its own `- run:` — so the local gate was strictly weaker
than the remote one, which is the opposite of its stated purpose.
Six steps were affected. The worst was `ng build ssp --localize && ng build
behandelportal --localize`: a missing English translation in ssp — the exact
thing the second-locale gate exists to catch — could not fail the run.
The one `( cd backend && ... )` step is safe as-is and left alone: a subshell
propagates its own non-zero status, so errexit sees it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Also records what RB-11 turned up: BIO-012 was factually wrong that the
proefbrief error mapping was already a separate function (it was inlined in a
try/catch), and the step-up literal is still a literal, moved one layer up to
the only caller rather than eliminated — BIO-006(c) stays a production gate.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Documents the dotnet swagger tofile crash discovered by actually
running the affected command (not just trusting ci-local.sh's local
"passed" line, which turned out to mask this exact failure via a
set -e && short-circuit gotcha), its root cause, and the two follow-up
fixes.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
RB-09 registers StubIdentityProvider only under IsDevelopment() and
throws for Production. `dotnet swagger tofile` (npm run gen:api) loads
the same Program.cs through .NET's design-time HostFactoryResolver,
which executes the app's startup code (including the unconditional
app.Services.GetRequiredService<IIdentityProvider>() the identity
middleware already relied on) without ever setting
ASPNETCORE_ENVIRONMENT - so it now defaults to Production and crashes
(dotnet swagger tofile exited 134), breaking `npm run gen:api`
entirely, including the "api-client drift" job in
.github/workflows/ci.yml (a separate `- run:` step there, so this
would fail real CI even though ci-local.sh's `cmd1 && cmd2` step
shape happens to swallow a cmd1 failure under `set -e` and reports
"passed" - a separate, pre-existing script fragility, not touched
here).
Real usage is unaffected: dotnet run already sets
ASPNETCORE_ENVIRONMENT=Development via launchSettings.json, and
docker-compose.yml/docker-compose.prod.yml already set
Development/Production explicitly - gen:api's bare CLI invocation was
the one place with no environment variable at all. Set it to
Development inline, the same value launchSettings.json already uses
for the real app.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
ehoandClaude Opus 5