- CLAUDE.md Key Conventions: 'App bootstrap' rule rewritten as 'Three
binding modes per feature' — describes USE_DEV_SEED + NODE_ENV
resolution order and the new ./di/bind-dev-seed export.
- AGENTS.md (root): exports list now mentions ./ui + ./di/bind-dev-seed;
Per-feature public-API surface table gains a row; Apps section shows
the bindAll() dispatcher with three-rule logic.
- docs/architecture/vertical-feature-spec.md §6: file shape now
includes bind-dev-seed.ts, bind-dev-seed.test.ts, __seeds__/dev.ts;
package.json exports list updated to include ./di/bind-dev-seed.
- docs/architecture/data-flow-explainer.html: anatomy tree gains
__seeds__/ row; LAYERS.di description updated with new binders +
cross-link to di-explainer.html; new LAYERS.seeds entry; public-
surface card expanded to six subpaths.
- docs/superpowers/refactor-logs/2026-05-06-input-output-unification.md
§7: new 'Post-Plan-9: dev-seed binders' entry summarizing the rollout
(commits, per-feature additions, app wiring, tests, turbo, docs).
- bind-production.test.ts: dispatcher tests use vi.stubEnv (typesafe
way to test process.env in TypeScript 5+ with @types/node read-only
process.env types). 4 dispatcher tests + 2 bindAllProduction tests
= 7 tests total.
Three-rule resolution order in bindAll() (first match wins):
1. USE_DEV_SEED === 'true' → bindAllDevSeed (explicit override)
2. NODE_ENV === 'production' → bindAllProduction
3. otherwise → bindAllDevSeed (developer default)
Rationale: 'pnpm dev' should boot the app without requiring Payload to
be running locally — dev seed is the more useful default for non-
production environments. Production servers explicitly set
NODE_ENV=production and get the real binding. The USE_DEV_SEED override
remains the escape hatch (force seed in any NODE_ENV — e.g. staging
preview, design review).
bind-production.test.ts grows from 3 tests to 8 — covers the dispatcher
matrix:
- USE_DEV_SEED='true' wins even when NODE_ENV='production'
- NODE_ENV='production' (no override) → production
- NODE_ENV='development' → dev seed (default)
- NODE_ENV unset → dev seed (default)
- USE_DEV_SEED='false' treated as not-set (only the literal 'true' triggers)
- Pre-existing 'binds all five repos' test now also asserts bindProductionMedia
di-explainer.html conditions table + mode flag strings updated to match
the new three-rule logic.
Earlier commit 74b88f7 attempted this but the Edit failed silently
without a prior Read. ESLint's turbo/no-undeclared-env-vars rule was
still flagging the warning.
ESLint flagged that USE_DEV_SEED was read in apps/web-next without being
declared in root turbo.json globalEnv — added it.
DI explainer's conditions table previously listed 'Production server
(NODE_ENV=production)' as a scenario, but bindAll() only checks
USE_DEV_SEED — NODE_ENV is not read anywhere in source. Fixed the table
and the prod-mode flag string to match what the code actually does.
apps/web-next/src/server/bind-production.ts now exports three functions:
- bindAllProduction() — production-only binders
- bindAllDevSeed() — dev-seed-only binders (NEW, calls all 5 features)
- bindAll() — dispatcher that branches on USE_DEV_SEED env var
All page/route callers (page.tsx, about/page.tsx, blog/[slug]/page.tsx,
api/trpc/[trpc]/route.ts) updated from bindAllProduction → bindAll so the
env flag actually has effect.
docs/architecture/di-explainer.html (NEW): standalone interactive page
explaining the di/ folder file-by-file, the loading sequence (8 stages),
the three binding kinds (.to / .toDynamicValue / .toConstantValue), an
interactive three-mode picker showing how the same blogContainer state
differs across default/dev-seed/production, a conditions table, and a
final card on how tests bypass DI entirely. Sister page to
data-flow-explainer.html.
Refactor log entry + canonical doc updates follow in subsequent commits.
Mirrors the canonical blog pattern landed earlier on this branch.
Per feature:
- src/__seeds__/dev.ts — lazy buildDev<Entities>() function using the
feature's existing factory for sensible defaults
- src/di/bind-dev-seed.ts — bindDevSeed<Feature>() async function that
rebinds the repo symbol(s) to a populated MockXRepository via
.toConstantValue
- src/di/bind-dev-seed.test.ts — 3+ tests per feature (populates,
reachable by id/slug, idempotent)
- package.json — adds ./di/bind-dev-seed subpath export
Tests + use cases continue to construct mocks directly; the seed never
runs from a *.test.ts path. App boot wiring (USE_DEV_SEED env branch)
follows in a separate commit.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Sets the canonical pattern for all 5 features. Three new files:
- src/__seeds__/dev.ts — buildDevArticles() function returning 3
realistic articles (welcome, vertical-feature-architecture,
wip-post) built from articleFactory with id/slug/title/status
overrides only.
- src/di/bind-dev-seed.ts — bindDevSeedBlog() async function that
unbinds IArticlesRepository, constructs MockArticlesRepository,
seeds it via buildDevArticles(), and rebinds via .toConstantValue.
- src/di/bind-dev-seed.test.ts — 3 tests: populated repo, welcome
article reachable by slug, idempotent (callable twice).
package.json adds the ./di/bind-dev-seed subpath export, parallel to
./di/bind-production.
Tests + use cases continue to construct MockArticlesRepository
directly — they never go through bindDevSeedBlog. The seed only attaches
when called explicitly at app boot.
User pushback: 'I already need mock data for dev and for testing at
runtime' — i.e., the dual role of the mock is obvious, not surprising.
Earlier framing was condescending.
Reframe around 'runtime reach':
- Section blurb leads with 'three artifacts at different distances
from runtime'
- Three-roles card intro: same neighborhood, different reach. Mock is
reached by runtime; contract + factory only by tests.
- 'The mock has two jobs' → 'The mock is reached from two directions
— both legitimate, neither is the test version'.
- Contract framing now leads with its real purpose: it tests the mock
alongside the real impl so you can trust the mock as a runtime
artifact.
User asked: 'arent mocks also related to testing?' — yes, but the mock
plays two roles where the contract and factory play one each. §04 now
opens with that framing.
What's new:
- Section title bumped to 'Mocks, contracts & factories'
- New three-roles diagram at the top of §04 visualizing the chain:
IArticlesRepository → MockArticlesRepository + ArticlesRepository
→ articlesRepositoryContract → articleFactory
- Two ordered lists explaining the mock's dual job (DI default binding
+ direct test fake) and the contract/factory's single test-only roles
- Three new expandable code examples:
· the mock as DI binding (real BlogModule code)
· the mock as direct test fake (use-case test, no container)
· the contract running against both impls (proof-of-parity)
- New CSS for .three-roles-diagram, .role-* boxes, .role-arrow with
vertical connector lines, .role-jobs lists with counter-leading
decimal numbering
The existing Contracts and Factories cards stay in place below — they
provide the deeper detail on each individual artifact.
Single-file HTML at docs/architecture/data-flow-explainer.html.
Self-contained — Google Fonts (Fraunces + JetBrains Mono) is the only
external resource; all interactivity is vanilla JS, all diagrams are
inline SVG/CSS.
Five sections:
01. Feature anatomy — clickable folder tree, layer detail card swaps
to explain entities / application / infrastructure / interface-
adapters / di / integrations / ui / __factories__ / __contracts__
/ public surface.
02. Request flow — step-through pipeline (12 stages, including the
success/error fork) with prev/next/play and a feature picker
(auth, blog, marketing-pages, navigation, media). Each feature
swaps real code snippets, the procedures.ts error map, and the
use-case/controller pattern.
03. Dependency injection — interactive DI canvas with default/prod
toggle so the user sees the same symbol resolving to mock vs
real Payload-backed impl. Real module.ts and bind-production.ts
code blocks below.
04. Contracts and factories — expandable sections with the actual
defineContractSuite + defineFactory code from packages/blog.
05. Verdict — short answer (yes), tradeoffs, badge row.
Editorial aesthetic: cream paper, deep ink, oxblood accent, Fraunces
display + JetBrains Mono code. Subtle paper noise + ruled-line bg.
No CDN scripts beyond fonts.
Doc-pass agent flagged §9.4 still listed only '.', './cms', './api' as
the allowed public surface. ./ui was added in Plan 9 and
./di/bind-production in Plan 5; both belong in the canonical list.
§1 worked example rewritten to direct factory injection (no container
rebinding). §4 mocking decision tree updated. New §sections for R25
output-validation tests, R26 router error-mapping tests, R27/R28
presenter-shape tests. §9 contract-suite paths use the post-Plan-8
.mock.ts suffix.
Refactor log doc-update checklist: tdd-workflow.md ticked.
Each per-feature AGENTS.md now reflects the post-Plan-9 layout:
entity/error paths, public-API split (./ui), use-case schemas, presenter
pattern, feature-scoped tRPC error map, and feature-specific
errors-to-codes table.
core-testing/AGENTS.md gains a Plan 9 test-patterns section documenting
R25 (output validation), R26 (router error mapping), R27/R28
(presenter shape) test obligations.
auth: documents real PayloadUsersRepository + AuthenticationService and
the deferred session methods.
media: documents the full Clean Architecture scaffold introduced in
Plan 8.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Direct factory injection is now the default mocking pattern (Plan 8).
Container rebinding is reserved for router-level tests.
rebindRepository helpers removed — use cases and controllers are factory
functions; tests construct mocks and pass them in.
New 'Test obligations per layer (Plan 9)' table maps each layer to its
required test types: R10 (controller input), R25 (use-case output
validation), R26 (router error mapping), R27/R28 (presenter view shape).
Refactor log doc-update checklist: testing-strategy.md ticked.
First slice of the combined Plan 8 + Plan 9 doc-update pass:
- CLAUDE.md Key Conventions: append schema-in-use-case, presenter,
controller unknown input, feature-scoped tRPC error mapping, public
surface split (./ui)
- packages/core-shared/AGENTS.md: document defineErrorMiddleware export
+ t re-export from trpc/init
- docs/superpowers/plans/2026-05-05-plan-8-*.md and matching spec:
one-line note that some controller/router patterns shifted in Plan 9;
link to the Plan 9 refactor log
- docs/architecture/overview.md: data-flow box now shows xProcedure +
xInputSchema + xOutputSchema.parse + presenter + middleware lanes;
three explanatory paragraphs added (schemas, presenter, error mapping)
- docs/architecture/dependency-flow.md: app-side ./ui subpath note,
allowed/disallowed examples updated for Plan 9 paths
Remaining doc-pass items (root AGENTS.md, per-feature AGENTS.md ×5,
core-testing AGENTS.md, adding-a-feature.md, tdd-workflow.md,
testing-strategy.md, vertical-feature-spec.md) follow in subsequent
commits — to be dispatched in parallel.
Final reviewer flagged two changelog inaccuracies:
- Summary said 14 commits; actual is 15 (88db39b..HEAD).
- §7 Task 8 said +144 tests; correct delta is +35 (325 → 360, +11%).
The +144 figure was a typo from the implementer's draft.
Both reconciled. Plan 9 ships.
Records the Plan 9 architectural decision (schemas in use-case file,
runtime output validation, presenter pattern, feature-scoped error
middleware, ./ui subpath split). ADR-012 gets a one-line cross-
reference to the new ADR. Refactor log gets a Summary section with
commit table and conformance checklist.
Plan 9 complete. The deferred doc-update pass (CLAUDE.md / AGENTS.md /
guides) — combined with the still-pending Plan 8 items — is the next
follow-up.
Refactor log: Summary, doc-update checklist
Spec: R29, R30
Straggler fixes: web-next and web-tanstack app callers did not pass {}
to queryOptions()/caller calls after Plan 9 added .input(z.object({}).strict())
to siteSettings and header procedures. All 360 tests pass, full typecheck
green across 14 packages. Refactor log §7 updated with verification summary.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Spec reviewer caught a systemic gap during Task 6 (navigation) review:
Plan 8 left every feature's domain error class with a constructor that
didn't set this.name = '<ClassName>'. Plan 9 spec R6 requires it.
Fixed across all 10 error files:
- auth: AuthenticationError, UnauthenticatedError, UnauthorizedError, InputParseError
- blog: ArticleNotFoundError, InputParseError
- marketing-pages: PageNotFoundError, InputParseError
- navigation: HeaderNotFoundError, InputParseError
- media: MediaNotFoundError, InputParseError
Functionally a no-op — defineErrorMiddleware uses instanceof, not name —
but ensures correct serialization, stack traces, and JSON inspection.
Refactor log: §7
Spec: R6
Code-quality reviewer flagged stylistic divergence from auth's pattern:
- All 3 blog use cases now have // ── Input ──... / ── Output ──... /
── Use case ──... section dividers (consistent with auth Task 3 + plan
template §4.1).
- Removed inline comment in get-articles controller's presenter that
appeared in only 1 of 3 controllers; identity presenters are
self-explanatory.
Code-quality reviewer flagged:
- R25 'malformed output' tests used .rejects.toThrow(/parse|invalid/i)
which could match unrelated errors. Replaced with
.rejects.toBeInstanceOf(ZodError) for both sign-in and sign-up.
- sign-in controller test asserted result.name only as truthy; the
parallel sign-up test was already precise (toBe('session')). Tightened
to match.
61 tests still passing for @repo/auth.
Spec reviewer noted §6.3 was missed during Task 3 commit. The presenter
tests exist in code (sign-in / sign-up controller tests assert cookie
shape, not full use-case output); the changelog just didn't record it.
Pure docs change.
Per Plan 9 (spec R1-R28):
- Use cases: input + output schemas (signIn, signUp); input-only for
signOut (void output). Use case body validates output via
outputSchema.parse before returning.
- Controllers: receive `unknown`; safeParse with the use-case schema;
presenter (returning cookie) for signIn/signUp; void return for
signOut.
- New integrations/api/procedures.ts with authProcedure built via
defineErrorMiddleware([[InputParseError,"BAD_REQUEST"],
[AuthenticationError,"UNAUTHORIZED"], [UnauthenticatedError,
"UNAUTHORIZED"], [UnauthorizedError,"FORBIDDEN"]]).
- Router uses authProcedure + .input(xInputSchema) for every procedure.
- src/index.ts exports schemas + types + IUseCase/IController aliases.
- package.json gains ./ui subpath; src/ui/index.ts placeholder
(auth has no query builders today).
- New tests: R25 output-validation per use case (signIn, signUp);
R26 router error-mapping (UNAUTHORIZED on missing user,
BAD_REQUEST on schema fail).
Refactor log: §1, §2, §3.1, §3.2, §3.3, §5.1, §5.2, §6.1, §6.2
Spec: R1–R6, R8–R15, R18, R19, R22–R26
Code-quality reviewer flagged a stale 'try/catch' phrase in the jsdoc
that didn't match the actual result.ok inspection pattern. Updated
description to accurately describe how tRPC v11 middleware sees errors
(returned result, not thrown exception) and how the middleware
discriminates and rethrows.
Pure docstring change — no behavioral or signature impact.
Factory takes [[ErrorCtor, TRPC_CODE], ...] tuples and returns a tRPC
middleware that translates matching domain errors to TRPCError. Discrim-
inates by instanceof; preserves original error as cause; unmapped
errors propagate (tRPC then wraps them as INTERNAL_SERVER_ERROR with
the original error as .cause — middleware does not interfere).
core-shared never enumerates feature errors — each feature passes its
own constructors in via integrations/api/procedures.ts (Tasks 3-7).
Also exports the `t` instance from trpc/init.ts so feature procedure
files can do t.procedure.use(...).
Also fixes tsconfig.json: rootDir set to "." and @/* path alias added
so test files using @/ resolve correctly under tsc --noEmit.
Refactor log: §1, §2, §4
Spec: R13–R17
Empty section template plus the doc-update checklist that the
post-Plan-9 follow-up pass will work through (combined with the still-
pending Plan 8 items so docs are written once).
Spec: docs/superpowers/specs/2026-05-06-input-output-unification-design.md §10, R30.
9 tasks, TDD throughout, one commit per task:
1. Refactor changelog scaffold
2. core-shared/trpc/define-error-middleware.ts factory + tests
3. auth migration (3 use cases — schemas + presenter + procedures.ts + router + ./ui + tests)
4. blog migration (3 use cases)
5. marketing-pages migration (2 use cases)
6. navigation migration (1 use case)
7. media migration (3 use cases)
8. Final verification + boundary sweep
9. ADR-013 + final changelog summary
Every step has concrete code; no placeholders. Self-review shows full
spec coverage (R1–R30) with each rule mapped to one or more tasks.
Spec: docs/superpowers/specs/2026-05-06-input-output-unification-design.md
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Codifies a single source of truth for use-case input/output contracts:
- Every use case exports xInputSchema/XInput and xOutputSchema/XOutput
from its own file; controllers and tRPC procedures import the same
schema. Use case body runtime-validates output via .parse before
returning.
- Controllers gain a co-located top-level `function presenter`
(Lazar-style) that reshapes entity output to view DTO; controller
return type = ReturnType<typeof presenter>.
- Domain error → TRPCError mapping moves to per-feature
integrations/api/procedures.ts via a defineErrorMiddleware factory
in core-shared/trpc/. core-shared never enumerates feature errors.
- Per-feature public-API surface cleaned: feature root = contracts
only (types, errors, schemas, IUseCase aliases, router type); new
./ui subpath holds query builders / components.
30 numbered rules using RFC-2119 MUST/SHOULD language so every
enforceable decision is citable from CLAUDE.md and per-feature AGENTS.md
during the post-Plan-9 doc-update pass. Includes file-shape templates,
migration order (~9 commits), doc-update mapping per rule, and an
ADR-013 mandate.
Spec: docs/superpowers/specs/2026-05-06-input-output-unification-design.md
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
First slice of the Plan 8 deferred doc-update checklist:
- CLAUDE.md Key Conventions: factory-function use cases/controllers,
entities/models/<x>.ts paths, .toDynamicValue DI bindings, direct
injection in tests
- docs/architecture/overview.md data-flow box updated to factory style
(controller resolved via container.get<IXController>; use case factory
takes deps as args)
- docs/decisions/adr-012-lazar-conformance.md created — records the
conformance decision and four intentional divergences
- docs/superpowers/plans/2026-05-05-plan-7-tdd-foundation.md and the
matching spec annotated with a "pre-Plan-8 layout" note pointing at
the refactor log
Remaining Plan 8 doc-update items (root AGENTS.md, per-feature AGENTS.md,
adding-a-feature.md, tdd-workflow.md, testing-strategy.md,
vertical-feature-spec.md §6/§10, core-testing AGENTS.md) intentionally
paused — Plan 9 (input/output unification) will change overlapping
content, so resuming after Plan 9 lands avoids double-churn.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
All conformance checks pass per spec §12 acceptance criteria. Summary
section populated with 9 task commits, file count breakdown, and
verification results. Doc-update checklist (deferred) remains untouched
— ready for follow-up doc pass.
Test count: 244 → 325 (+81). All features now structurally conformant
with Lazar Nikolov Clean Architecture pattern (with intentional
divergences documented in spec §4).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Use case (get-header) → factory function with IGetHeaderUseCase alias
- Controller renamed header.controller.ts → get-header.controller.ts (verb-noun); converted to factory function with IGetHeaderController alias
- DI module wires factories with .toDynamicValue()
- tRPC router resolves controller via container
- Use case + controller tests refactored to direct factory injection (no container rebinding)
- container.test.ts verifies IGetHeaderUseCase + IGetHeaderController symbols
Refactor log: §1, §4.1, §4.2, §5.1
Spec: §6.4
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Use cases (sign-in, sign-up, sign-out) → factory functions with I*UseCase aliases
- Controllers → factory functions with I*Controller aliases
- DI symbols + module updated with .toDynamicValue() bindings for factories
- New: real UsersRepository (Payload-backed, SanitizedConfig, contract-tested)
- New: real AuthenticationService (node:crypto hashing/UUIDs; createSession/
validateSession/invalidateSession deferred — see refactor log §7)
- bindProductionAuth swaps both mocks for real impls (was a no-op before)
- Tests refactored to construct mocks and inject directly (no container rebinding)
- Feature test constructs full chain via direct factory injection
Refactor log: §2, §4.1, §4.2, §5.1, §5.2, §6.1, §7
Spec: §6.1, §7
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
All 5 features (auth, blog, marketing-pages, navigation; media has no
entities yet) now follow Lazar's pattern:
- entities/<x>.ts → entities/models/<x>.ts
- entities/errors.ts → entities/errors/<domain>.ts + errors/common.ts
Updates all import paths across factories, contracts, tests, use cases,
controllers, repositories, integrations, and src/index.ts exports.
navigation divergence: had no errors.ts; errors/header.ts +
errors/common.ts added as new forward-looking stubs.
Refactor log: docs/superpowers/refactor-logs/2026-05-05-lazar-pattern-conformance.md
Spec: §5, §9.3
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Empty section template plus the full doc-update checklist that the
follow-up pass will work through after the refactor is merged. Captures
the substitution map (paths, naming, patterns) so the doc updater can
apply changes mechanically.
Spec: docs/superpowers/specs/2026-05-05-lazar-pattern-conformance-design.md §10
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Ten TDD'd tasks executing the Lazar Nikolov Clean Architecture pattern
across all 5 features. Task 1 scaffolds the refactor changelog (sole
artifact for the deferred doc-update pass). Tasks 2-3 are foundation
(entities split, file renames). Tasks 4-7 refactor each existing feature
to factory-function pattern. Task 8 scaffolds media as a full Clean
Architecture feature. Task 9 aligns factory/contract imports. Task 10
final verification.
External docs (CLAUDE.md, AGENTS.md, adding-a-feature.md, tdd-workflow.md,
ADR-012) explicitly NOT touched during the refactor — captured as a
checklist in the refactor changelog for a single batched doc-update pass
afterwards.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Brings every feature into structural conformance with the reference
nextjs-clean-architecture repo: factory-function use cases and
controllers, entities/{models,errors}/ split, .mock.ts file suffix,
per-use-case controller files, real PayloadUsersRepository and
PayloadAuthenticationService for auth, full Clean Architecture
scaffold for media. Documents intentional divergences (per-feature DI,
inversify retained, colocated tests). All external doc updates
deferred to a follow-up pass driven by the refactor changelog.
Spec: docs/superpowers/specs/2026-05-05-lazar-pattern-conformance-design.md
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
