Extract ScannedCardItem with per-card metadata controls and ownership
lookup via GET /api/cards/[id]/ownership. Propagate condition, foil,
and quantity through owned/collection/deck POST paths.
Co-authored-by: Cursor <cursoragent@cursor.com>
Add ScannerDestinationPicker with game filter and owned/collection/deck
targets. Persist session destination in localStorage and auto-add each
identified scan to the active destination.
Co-authored-by: Cursor <cursoragent@cursor.com>
Add per-row in-flight locks so double-tap cannot duplicate owned POSTs.
Pass bulk action/target directly instead of setTimeout state races.
Log collection card adds via logCollectionActivity and fix rows.length
checks in the collection cards POST handler.
Co-authored-by: Cursor <cursoragent@cursor.com>
When vision reads a set+number missing from the catalog, route to
card_submissions rather than sibling disambiguation. Adds a not-listed
modal action, background vision refine, foil-friendly prompt, and
submit-for-review API. Queues catalog-sync-vercel-cron convoy for later.
Co-authored-by: Cursor <cursoragent@cursor.com>
Fix identify-by-text 500 (Neon could not infer null game param type).
When a card name has multiple catalog printings, show disambiguation
instead of auto-picking the first match. Throttle concurrent vision
calls and suppress repeated 429/error toasts during detection.
Co-authored-by: Cursor <cursoragent@cursor.com>
Route Layer-2 identification through Vercel AI Gateway (AI_GATEWAY_API_KEY,
default google/gemini-2.5-flash-lite). Add Layer-1 browser Tesseract name-strip
OCR with pg_trgm fuzzy catalog match via /api/cards/identify-by-text before
escalating to vision.
Co-authored-by: Cursor <cursoragent@cursor.com>
Use the same vision model as the deleted browser client, surface Gemini
quota/denial/migration failures as 502/503 with readable text, and stop
scan_attempts telemetry from blocking identification.
Co-authored-by: Cursor <cursoragent@cursor.com>
Scanner was sending logged-in users to /login while useAuth was still
loading. Admin card-editor/card-import crashed on login because hooks
ran after a mounted early return (Rules of Hooks violation).
Co-authored-by: Cursor <cursoragent@cursor.com>
* feat(scanner): move card identification server-side (convoy #2)
Replace browser Gemini/OCR with POST /api/scan/identify, add card_submissions
review queue, remove user-writable cards INSERT, and surface disambiguation
when catalog matching is ambiguous.
Co-authored-by: Cursor <cursoragent@cursor.com>
* ci: allowlist server-only lib/scan-gemini.js in LLM key gate
The scan pipeline helper lives under lib/ but is imported exclusively
from pages/api/scan/identify — exclude it from the client-side URL scan.
Co-authored-by: Cursor <cursoragent@cursor.com>
---------
Co-authored-by: Cursor <cursoragent@cursor.com>
Delete the public /api/config/gemini endpoint and remove client auto-load
paths so GEMINI_AI_API_KEY stays server-side only. Add a scan rate-limit
class for the upcoming server-side identify route and a CI gate that blocks
reintroducing config key leaks or new browser LLM URLs.
Co-authored-by: Cursor <cursoragent@cursor.com>
Closes P1 #11 of .convoys/ship-readiness.md (launch sequence step 7) —
"No migration tool — scripts/add-*.js graveyard". Schema changes
post-this-convoy ship as node-pg-migrate migrations under migrations/
at the repo root; the legacy 27 scripts/add-*.js / scripts/fix-*.js /
scripts/seed-*.js jobs remain append-only history per the no-go-zones
rule.
Decisions (full record in .convoys/migration-tool.md § Decisions):
D1 — Tool: node-pg-migrate@^8. Rejected drizzle-kit / prisma migrate /
kysely because each forces broader TypeScript surface than AGENTS.md
Gotcha #9 allows (TS is a devDep only). node-pg-migrate is
JavaScript-native, raw-SQL-friendly via pgm.sql(), and ESM-clean for
the post-bump-next-js "type": "module" repo. Brings pg@^8.21.0 as a
peer dep (dev-only; never loaded in the Next.js bundle).
D2 — Migrations directory: migrations/ at the repo root. Separates
the tool-wrapped artifacts from the historical scripts/migrations/
placeholder folder (which housed the lone pre-tool
2026-05-24-rename-admin-email.js migration and remains preserved for
the audit trail). Matches node-pg-migrate's default flag.
D3 — Tracking table: default pgmigrations (no name collision with
the existing 7-table bootstrap; zero CLI noise).
D4 — Backfill strategy: hand-translate scripts/setup-neon-db.js's
DDL into the initial migration verbatim. Each await sql`...` block
becomes one pgm.sql(`...`) call. Each CREATE uses IF NOT EXISTS, so
the migration is idempotent against fresh AND pre-existing envs —
re-running setup-db on an env that already has the schema is a no-op
DDL-wise (only records the pgmigrations row). Documented assumption:
prod has drifted via the 27 historical add-*.js scripts; reconciling
those into the migration history is the queued
reconcile-historical-add-scripts follow-up convoy.
D5 — Bootstrap reconciliation: split. setup-neon-db.js now (1)
validates ADMIN_INITIAL_PASSWORD + POSTGRES_URL, (2) spawns
`npm run migrate up` via child_process with stdio inherited, (3)
seeds the admin row with ON CONFLICT (email) DO NOTHING. The seven
DDL blocks are deleted from setup-neon-db.js; success/error message
copy is updated to mention the migration step explicitly.
D6 — CI integration: defer. Wiring a CI job that runs migrate up
against a test DB needs either a dedicated Neon branch + secret OR a
Postgres service container; both are real work. Surface as
wire-migrate-into-ci follow-up. Risk acknowledged in
.convoys/migration-tool.md § R3.
D7 — Down-migration on the initial backfill: hard stub. Rolling back
the initial schema would drop every user / card / collection / deck
row in the DB. The stub throws with a long-form error pointing at
the recommended alternative (branch the Neon database + forward-apply).
Future migrations that touch one of the seven bootstrap tables write
their own dated migration with a real down().
Verification (pre-PR):
- npm run lint → 128 problems (baseline preserved, zero regression;
migration file is lint-clean, no new ignore patterns)
- npm run test:run → 21/21 pass
- node --check on migrations/1779853647564_initial-schema.js + on
scripts/setup-neon-db.js → exit 0
- Module load + down() throw verified via dynamic import
- npm run migrate -- --help reaches the node-pg-migrate CLI through
the wrapper
Live verification against a Neon branch is deferred (no throwaway
branch available); the operator's optional post-merge sequence is
documented in .convoys/migration-tool.md § Operator runbook.
See .convoys/migration-tool.md § Follow-ups for the queued
wire-migrate-into-ci / reconcile-historical-add-scripts /
retire-graveyard-scripts-after-audit / audit-node-pg-migrate-transitive-deps
/ add-migration-template follow-up convoys.
Co-authored-by: Cursor <cursoragent@cursor.com>
`lib/use-auth.js` is now the sole client-side auth surface (P1 §9 of
`.convoys/ship-readiness.md`). The legacy `lib/auth-context.js`
(`AuthProvider` + `useAuth`) and `lib/admin-auth.js` (`AdminProvider` +
`useAdmin` + `useIsAdmin`) are deleted; every importer is migrated to
the canonical hook. Pre-convoy a worst-case page mount issued THREE
identical `GET /api/auth/verify` requests (one per provider/hook); the
post-convoy floor is one verify per page mount (3 → 1 on
`pages/card/[id].js`, 2 → 1 elsewhere).
Importer inventory swept (7 source files):
- `pages/_app.js` — removed `<AuthProvider>` wrapper; `<ThemeProvider>`
is now the only top-level provider. `lib/use-auth.js` is hook-only,
no replacement provider needed.
- `pages/index.js`, `pages/scanner.js`, `pages/decks.js`,
`pages/deck/[id].js`, `pages/deck-builder.js` — `import { useAuth }`
path swap from `../lib/auth-context` to `../lib/use-auth`. All five
pages destructured only `{ user }` or `{ user, loading }`; verified
no consumer reads `login` / `register` from useAuth (those flows are
in `pages/login.js` / `pages/signup.js` which call the API directly),
so no shape-parity gap on `lib/use-auth.js`.
- `pages/card/[id].js` — replaced `useIsAdmin()` (the only consumer of
`lib/admin-auth.js` anywhere in the tree) with synchronous
`user?.role === 'admin'` derived from the existing `useAuth()` call.
Render condition at line 524 stays byte-identical.
Decisions documented in `.convoys/single-auth-provider.md`:
- D1: no extension to `lib/use-auth.js` (zero call sites for `login` /
`register` from useAuth — those flows are direct fetches in
`login.js` / `signup.js`).
- D2: `useIsAdmin()` collapses onto `useAuth()`; no separate hook.
- D3: provider tree `<ThemeProvider><AuthProvider>{children}</AuthProvider></ThemeProvider>`
→ `<ThemeProvider>{children}</ThemeProvider>`.
- D4: 3 → 1 verify roundtrip on `card/[id].js`; 2 → 1 on every other
page-load.
- D5: zero test files modified; the 21-test vitest suite is server-
side or prop-driven (`Layout.test.js` passes `user` as a prop, never
imports the legacy hooks).
Doc / config updates so the deletion lands cleanly:
- `.github/CODEOWNERS` — drop the two CODEOWNERS lines for the deleted
files.
- `AGENTS.md` § 2 architecture row + § 3 "Auth (client)" bullet —
rewritten for the post-convoy single-surface state.
- `.cursor/rules/auth-and-permissions.mdc` — § "Legacy" reframed to
"deleted by this convoy"; § "Authentication state on the client"
updated to the post-convoy `useAuth()` shape and the direct-fetch
login flow used by `login.js` / `signup.js`.
- `.cursor/rules/no-go-zones.mdc` — auth-refactors bullet drops the
deleted files from the canonical list.
- `.cursor/skills/add-page/SKILL.md` — checklist + anti-pattern row
refer to the deletion.
Verification:
- `rg "lib/auth-context|lib/admin-auth" --type js` → 0 hits in source.
- `npm run lint` → 128 → 125 problems (3 fewer errors from the deleted
unused-import lines; no regression).
- `npm run test:run` → 21/21 pass (including the 5 Layout regression
locks from `fix-layout-default-user`, which are prop-driven and
unaffected).
- `npm run build` → all 26 pages compile end-to-end; no SSR / static-
generation breakage that would have surfaced if a page tried to use
the legacy context hook unwrapped.
- Manual smoke deferred to operator post-merge per convoy doc.
Risks (full discussion in convoy file):
- R1 shape parity gap — verified zero consumers of legacy-only
surface; mitigated.
- R2 SSR mismatch from removing `<AuthProvider>` — `useEffect`-
guarded `localStorage` read; identical SSR shape pre/post; build
passes.
- R3 missed importer — post-delete grep + build pass would surface
any miss.
- R5 stale `useAuth` cache across components — pre-existing
pattern, called out as follow-up rather than addressed here.
Out of scope: any change to `lib/permission-middleware.js` (server-
side; resolved P0 #1), `lib/auth-secret.js` (resolved P0 #2),
`pages/api/**` route handlers, login / register API contracts, or
the seeded admin account flow.
Co-authored-by: Cursor <cursoragent@cursor.com>
Convoy: single-sql-client (P1 quality, launch sequence step 8)
Addresses: AGENTS.md Gotcha #1, .convoys/ship-readiness.md P1 #8
## Decisions
- D1: Caller inventory = 2 files (1 source + 1 test), not "~3 based on graph".
Only pages/api/auth-utils.js imports `db`; test/api/auth-utils.test.js mocks
it purely to satisfy the import graph (the 5 tests exercise
generateToken/verifyToken, not isAdmin/getUserById).
- D2: Migrate both call sites (isAdmin, getUserById) to @vercel/postgres
tagged-template SQL. Queries are SELECT-only, single-table,
single-numeric-parameter — byte-equivalent translation; same result shape
({rows, rowCount}); no transaction or pool semantics differ.
- D3: KEEP @neondatabase/serverless as a dep. 11 scripts/* files still use
`neon()` directly (setup-neon-db.js, migrations/, reset-db.js, 8 historical
add-*/fix-*/seed-* jobs). They are out of scope per the no-go-zones rule
and the convoy spec; purging the dep entirely would be its own convoy
(queued as `purge-neondatabase-serverless-fully`, blocked on migration-tool).
- D4: sql.unsafe audit — NOT a real injection vector with current callers
(userId comes from a verified JWT, is a numeric SERIAL id). Security
finding: NO. Pure refactor + foot-gun removal that prevents the FUTURE
caller that would have been the incident.
- D5: Test mock cleanup — drop the now-unneeded `vi.mock('../../lib/database.js')`
call + unused `vi` import. Test count + assertions unchanged (5/5).
## Per-file changes
- pages/api/auth-utils.js: swap `import { db } from '../../lib/database.js'`
for `import { sql } from '@vercel/postgres'`; rewrite isAdmin's
`db.query(SELECT … WHERE id = $1, [userId])` and getUserById's same shape
to `sql\`SELECT … WHERE id = ${userId}\``. Same try/catch, same
result.rows[0] access, same error returns.
- test/api/auth-utils.test.js: drop vi.mock for lib/database.js + the unused
`vi` import. 5/5 tests still pass.
- lib/database.js: DELETED (47 lines removed; manual-interpolation + sql.unsafe
wrapper is gone).
- .convoys/single-sql-client.md: NEW (the convoy file documenting all
decisions + caller inventory + verification + risks + follow-ups).
## Verification
- npm run lint → 128 problems (baseline preserved, no regression)
- npm run test:run → 21/21 pass (vitest)
- Grep "lib/database" --type js -l → 0 hits anywhere
- Grep "@neondatabase/serverless" --type js -l → still matches the 11
scripts/* sites (expected; out of scope per D3)
- node --check pages/api/auth-utils.js → exit 0
## Scope note
This convoy collapses the lib/database.js abstraction onto the canonical
@vercel/postgres surface for pages/api/**. It does NOT eliminate
@neondatabase/serverless from the dependency tree — that would require
migrating the scripts/* helpers, which is out of scope here (no-go-zones
rule + convoy spec). Queued as a follow-up.
## Live smoke
Deferred. The two migrated functions (isAdmin, getUserById) are only
reachable via pages/api/admin/index.js which requires an admin Bearer
token and a populated users table in prod Neon. Byte-equivalent SQL +
identical result shape gives high confidence; rollback is a single-commit
revert if a post-merge admin action 500s.
Co-authored-by: Cursor <cursoragent@cursor.com>
Add an ESLint no-restricted-syntax rule scoped to scripts/**/*.js
that flags any CallExpression with callee name `require`. The
recurring bug pattern: helper scripts under scripts/ that use CJS
require() throw `ReferenceError: require is not defined` on Node
22.x because package.json has had "type": "module" since
bump-next-js. The bug has bitten twice in two convoys — once in
drop-public-setup Brief 2 (setup-neon-db.js, commit b63b509) and
again in fix-reset-db-script (reset-db.js, PR #25 squash 3ab9bf8).
Both were caught at first run, not at lint time. This rule would
have caught both at PR time.
Rule shape: a second flat-config block at the end of
eslint.config.mjs (NOT in the root rules block) targeting only
scripts/**/*.js. The error message points at
.convoys/fix-reset-db-script.md so the next agent who trips it
gets a 1-click path to the exemplar fix (ESM top-level imports
for dotenv, neon, bcrypt) instead of having to re-derive it.
scripts/migrations/** is already in globalIgnores from pick-a-name
Brief 2 and stays excluded.
Blast-radius rationale (scripts/** only, not all .js at repo root):
matches the actual observed bug surface. pages/api/** is already
correctly ESM-imported throughout (verified across add-rate-limiting,
cors-tighten, and the add-route skill). The config files
(postcss.config.js, tailwind.config.js, next.config.js) intentionally
use CJS-style exports that the next-config base rules already handle
correctly. A repo-wide ban would produce zero true positives outside
scripts/** today and would require explicit allowlist for every
config file — strictly more code, more maintenance, zero benefit.
Convoy file: .convoys/lint-against-cjs-in-esm-scripts.md (P3 polish,
parent-owned, no architect — preventative one-line rule following
two proven bug recurrences).
Verification:
- node --check eslint.config.mjs: exit 0
- npm run lint: 128 problems (81 errors, 47 warnings) — baseline
preserved verbatim, zero new false positives in current tree
- Negative test: prepended `const x = require('fs');` to
scripts/reset-db.js, ran npm run lint, observed exit 1 with
129 problems and the rule firing at line 20:11 with the
documented message, then reverted to 128 problems clean
- npm run test:run: 21/21 pass (no test surface touched)
- Grep: 0 require( occurrences in scripts/**/*.js (current tree
is clean; rule starts with zero positives to silence on day 1)
Surfaces no new follow-up — this convoy IS the follow-up surfaced
by fix-reset-db-script.
Co-authored-by: Cursor <cursoragent@cursor.com>
components/MobileNavigation.js has accepted a `user` prop ever since
the mobile bottom-bar was extracted from Layout, but it has never read
any field of `user`. The bottom-bar items (Cards, Decks, Dashboard,
Community, More) are statically configured — none of them branch on
auth state, role, user id, or any other per-user attribute. The prop
is dead.
This was originally surfaced as R8 in the fix-layout-default-user
convoy (commit ca302a8) and deliberately deferred there to keep that
convoy focused on the Layout default-user fix. The follow-up was
queued as cleanup-mobile-nav-dead-props in .convoys/ship-readiness.md
§ Queued convoys. Pre-edit audit confirms the queue entry's premise:
`rg '\\buser\\b' components/MobileNavigation.js` returns 1 hit (the
destructure on line 5) before the change and 0 hits after. The
only active call site is components/Layout.js line 598; the
components/Layout.js.backup snapshot also calls it but is a
no-go-zone (per .cursor/rules/no-go-zones.mdc § "Append-only /
historical") and stays untouched — when that backup is eventually
deleted in a separate convoy, its stale call disappears with it.
Verification: npm run lint exit 1 with 128 problems (baseline
preserved, no regression introduced); npm run test:run 21/21 pass
(test/components/Layout.test.js still asserts the logged-out branch
contract from PR #15 — the dead-prop removal is invisible to that
suite since it does not inspect MobileNavigation's prop shape).
Convoy file .convoys/cleanup-mobile-nav-dead-props.md captures the
audit, fix, risks (R1: a future per-user bottom-bar feature would
need to re-add the prop — accepted; carrying dead state to hedge
hypothetical features is worse than paying the one-line re-add cost
when the feature actually lands), and acceptance criteria.
Co-authored-by: Cursor <cursoragent@cursor.com>
`scripts/create-test-users.js` hardcoded `bcrypt.hash('alice123', 12)`
+ `bcrypt.hash('bob123', 12)` and echoed those literals back to stdout
both per-user and in a final summary block. `TESTING_GUIDE.md`'s Test
Accounts table documented the same `admin123` / `alice123` / `bob123`
trio. These were the last two weak-credential surfaces left in the
helper-script + manual-QA-doc tree after `drop-public-setup` (commits
`ff80753` + `b63b509`) and `fix-reset-db-script` (squash `3ab9bf8`,
PR #25) closed the `setup-neon-db.js` and `reset-db.js` halves of the
umbrella `purge-weak-creds-from-helpers` queued follow-up.
The fix mirrors the post-`drop-public-setup` `setup-neon-db.js`
pattern and the post-PR-#25 `reset-db.js` pattern verbatim, with one
deliberate simplification: a single `TEST_USERS_PASSWORD` env var
covers both alice + bob rather than per-user env vars (risk R2 in the
convoy file argues this — these are fixture users for the
collaboration demo flow, not independent identities, and per-user
sprawl would double the env-var contract for zero security benefit).
`createTestUsers()` now reads `process.env.TEST_USERS_PASSWORD` at the
top of the function body and exits with code 1 BEFORE opening any DB
connection if the var is unset or whitespace-only, with the same
helpful-error wording template the other two scripts use (names the
var, points at `.env.local`, suggests `openssl rand -base64 24`,
references README's "First-time admin setup" section). All four
password-echo `console.log` lines are deleted; the new summary
documents *where* the password comes from without ever printing it.
`TESTING_GUIDE.md`'s Test Accounts table is rewritten to show password
source per user instead of the literal value; the two inline
`Password: alice123` / `Password: bob123` workflow snippets are
replaced with placeholder text. Unlike the previous two convoys, no
CJS→ESM conversion was needed — `create-test-users.js` was already
top-level ESM.
Verification (all static — script is destructive and not live-tested):
`node --check scripts/create-test-users.js` exit 0; `npm run lint` 128
problems (baseline preserved, no regression); `npm run test:run` 21/21
pass; grep `scripts/ TESTING_GUIDE.md` for
`admin123|password123|test123|alice123|bob123` → 0 hits;
`TEST_USERS_PASSWORD` referenced 10 times total (5 script + 5 doc).
Operator caveat: anyone running `node scripts/create-test-users.js`
post-merge must add `TEST_USERS_PASSWORD=<value>` to their
`.env.local` first; existing alice + bob rows in already-seeded
environments are NOT rotated by re-running this script
(`ON CONFLICT (email) DO NOTHING` preserves the old hashes). Same
caveat that applies to the `drop-public-setup` admin row.
Co-authored-by: Cursor <cursoragent@cursor.com>
The `Screenshot diff` workflow's `paths:` filter included `pages/**`
which matches `pages/api/**` too, so API-only PRs triggered the
visual-diff workflow even though they can't possibly move a single
rendered pixel. PR #19 (cors-tighten) and PR #20 (add-rate-limiting)
both empirically hit this — each was an API-only sweep, and each
burned ~55s of CI runtime on a `Screenshot diff` job that
`continue-on-error: true` then swallowed. Documented as a queued
follow-up in `.convoys/ship-readiness.md` § Queued convoys, ratified
for fix in this convoy.
The fix is a single negated-glob entry inserted immediately after
`pages/**` in the `paths:` list. GitHub Actions evaluates `paths:`
with minimatch and supports `!`-prefixed exclusions per the
published path-filter cheatsheet, but the order matters: a `!pattern`
only takes effect if it appears AFTER an include that already
matched the path. Keeping `!pages/api/**` second in the list (right
after `pages/**`, before all the other includes) is the canonical
shape. All five existing entries are preserved verbatim; only the
one exclusion entry plus an inline comment explaining the ordering
rule and the empirical motivation is added.
`preview-smoke.yml` is intentionally untouched — verified its `on:`
block has no `paths:` filter at all (it triggers on every PR
targeting main, with skip-via-PR-body-directive in the gate job),
so there's no false-positive shape to fix there. Smoke SHOULD run on
every PR including API-only ones because changes to `pages/api/**`
can break the home redirect + sign-in + `/api/health` endpoints the
smoke spec exercises.
Co-authored-by: Cursor <cursoragent@cursor.com>
Fill in fix-reset-db-script.md § As-shipped with the merged-state record
(PR #25, squash 3ab9bf8, +161/-18) — three bugs atomically resolved by
mirroring setup-neon-db.js post-drop-public-setup shape; CI green
including 5th-consecutive smoke-spec cross-validation; lint-against-
cjs-in-esm-scripts surfaced as new P3 queue entry.
Update ship-readiness.md Queued convoys:
- convert-reset-db-to-esm → RESOLVED (audit-trail kept, not removed)
- purge-weak-creds-from-helpers → scope reduced (reset-db.js half done;
create-test-users.js + TESTING_GUIDE.md remain)
- lint-against-cjs-in-esm-scripts → NEW (P3 polish; would have caught
both DPS-B2 and PR #25 at lint time)
Co-authored-by: Cursor <cursoragent@cursor.com>
Fold of two queued follow-ups from pick-a-name architect audit
(convert-reset-db-to-esm + purge-weak-creds-from-helpers). Three bugs
in one file; all three fixed atomically by mirroring the proven post-
drop-public-setup setup-neon-db.js shape (commit b63b509).
Bugs fixed:
1. CJS-in-ESM (lines 10, 12, 142): require('dotenv'), require('@neon...'),
inline require('bcryptjs'). package.json has "type": "module" since
bump-next-js, so npm run reset-db threw ReferenceError on Node 22.x.
Same bug pattern that hit setup-neon-db.js pre-drop-public-setup B2.
2. Hardcoded weak admin password (line 143: bcrypt.hash('admin123', 12)).
Same anti-pattern drop-public-setup B1 removed from setup-neon-db.js.
3. Password echoed to stdout (line 156: console.log('Admin Password:
admin123')). Security anti-pattern; setup-neon-db.js post-DPS does
NOT echo passwords.
Fix shape (verbatim mirror of setup-neon-db.js):
- ESM top-level imports (dotenv, neon, bcrypt)
- Fail-loud ADMIN_INITIAL_PASSWORD env-var check at function top with
helpful error message pointing to README "First-time admin setup"
- bcrypt.hash(adminPassword, 12) instead of literal
- ON CONFLICT (email) DO NOTHING on INSERT (defensive against
double-run, matches setup-neon-db.js line 149)
- No password echo in success block; admin email logged for confirmation
- Updated docstring to flag DESTRUCTIVE + reference required env
Convoy file: .convoys/fix-reset-db-script.md (P2 hygiene, parent-owned,
no architect — this is a proven-pattern fold with no new decisions
to ratify).
Verification:
- node --check scripts/reset-db.js: exit 0
- npm run lint: 128 problems (baseline preserved, no regression)
- npm run test:run: 21/21 pass
- Grep: 0 require( | 0 admin123 | 0 'Admin Password' in scripts/reset-db.js
- Grep: 3 ADMIN_INITIAL_PASSWORD references (docstring, const, error msg)
NOT live-tested (script is destructive — drops all tables). Operator
can optionally run npm run reset-db against a non-prod Neon branch
post-merge to verify end-to-end.
Surfaces follow-up: lint-against-cjs-in-esm-scripts (P3 polish — add
ESLint rule to prevent any future require() in scripts/** under
"type": "module"). Surfaced for future convoy queue.
Co-authored-by: Cursor <cursoragent@cursor.com>
Hotfix to scripts/migrations/2026-05-24-rename-admin-email.js (shipped 2026-05-24 in pick-a-name PR #21). Script crashed on first invocation with 'TypeError: Cannot read properties of undefined (reading length)' at line 44. Root cause: architect designed against @vercel/postgres return shape { rows, rowCount } but the script uses @neondatabase/serverless's neon() tagged template which returns the rows array directly. AGENTS.md Gotcha #1 (two SQL clients in parallel) is exactly this kind of cross-contamination. Fix: drop the { rows: x } destructuring in all 3 sites + add a 4-line why comment block above the first site so the next migration author doesn't repeat. Verified hand-run against prod Neon DB: migrated 3 users (admin id=1, alice id=5, bob id=6) from @tcgvault.com to @deckhearth.com; idempotent re-run prints 'Nothing to migrate.' No data risk on the original crash — script exited at line 44 before reaching the UPDATE at line 54. PR #21 operator action item now complete in prod. Surfaces a P3 follow-up: add-neon-return-shape-rule (or fold into single-sql-client). All CI green: lint 128 baseline, vitest 21/21, Playwright smoke 3/3 in 1m2s, forbidden-cors-headers pass, forbidden-endpoints pass. PR #24, commit 98406fa.
Post-merge doc cleanup for PR #21. Resolves the AGENTS.md line-5 'Pick one before launch' branding question. Updates AGENTS.md top branding note + Gotcha #4 + Gotcha #12, adds full as-shipped to .convoys/pick-a-name.md, adds 4 queued-convoy entries to ship-readiness.md, adds pick-a-name attribution to .cursor/rules/api-routes.mdc. Pure docs (+357/-15 across 4 files).
Reflects the merged add-rate-limiting convoy (PR #20, squash commit
708ef45) in repo documentation. **This is the milestone cleanup** —
add-rate-limiting closed P0 #6 (No rate limiting anywhere), the LAST
open P0 ship-blocker. `.convoys/ship-readiness.md`'s § Status summary
flips from "7 of 8 RESOLVED; 1 remains" to **"8 of 8 RESOLVED.
Launch-readiness P0 checklist is empty."** One brief in the convoy:
Brief 1 shipped as planned with no scope expansions and no implementer
deviations from the verbatim spec; all six architect decisions ratified
verbatim at gate 1 (D1 operator-ratified Option A; D2-D6
architect-self-ratified).
.convoys/add-rate-limiting.md:
- frontmatter status: in-progress -> shipped (added shipped: 2026-05-24)
- new ## As-shipped section. Opens with the milestone language
pointing back at ship-readiness.md's flipped § Status summary.
Decisions section captures all 6 ratifications (D1 operator-
ratified Option A — Critical: WHY the atomic admin UI fix in
pages/admin/card-import.js was Decision 1's hidden coupling
requirement, since API gating alone would have broken every
"Import Cards" click; D2 hybrid named-limiter shape with
Map<className, Ratelimit> cache; D3 per-class table including
the two D3 tuning-evidence raises — search 30 -> 60/min because
ShareModal.handleSearch has no debounce so a 17-char email = 16
requests in <5s, and generate kept at 5/hour because DiceBear is
free not paid AI; D4 two-extractor shape with defensive THROW
on null/empty userId; D5 uniform 429 message; D6 no new vitest
or playwright specs deferred to fill-vitest-handler-coverage).
As-shipped surface broken into 4 layers (1 lib refactor + 6 route
gates + 1 atomic admin UI fix + 1 rule extension) mirroring the
cors-tighten cleanup's pattern-split shape. Empirical CI metrics
from post-merge run 26382185019 (Playwright smoke 59s 3/3 in
3.8s, forbidden-cors-headers pass, vitest 21/21, lint 128 baseline,
Screenshot diff continue-on-error swallow per Decision 4).
Cross-validation finding: smoke test 2 still passes against the
post-rate-limit preview — that's three convoys in a row (PR #15
Layout default-user, PR #19 CORS-tighten, PR #20 rate-limiting)
where the same 3-test smoke spec defended the auth surface
through sweeping changes. Operator-action-required: none. What
did NOT change audit trail.
.convoys/ship-readiness.md:
- § Status summary at the top flipped from 7/8 to 8/8 RESOLVED.
Header text updated to "Launch-readiness P0 checklist is empty."
P0 #6 row in the table flips from PARTIAL to RESOLVED with the
two-convoy lineage (fix-auth-bypass Brief 4 + add-rate-limiting).
Trailing paragraph rewritten as a milestone note: security gate
closed; remaining launch work is P1 quality bar + P2/P3 polish.
- P0 #6 entry flipped from PARTIAL to RESOLVED with the full
add-rate-limiting as-shipped block (8 sub-bullets covering the
lib refactor shape, the per-class table, the defensive THROW,
the three import routes' auth-gating, the atomic admin UI fix
and WHY, the rule extension, the 6 decisions, and the diff
breakdown). Brief 4's 2026-05-23 partial is preserved as the
prior as-shipped layer to maintain the audit trail.
- Launch sequence step 4 marked RESOLVED 2026-05-24 with the
squash commit + smoke metrics inline.
- Queued convoys: removed the add-rate-limiting entry (it shipped).
Added a new delete-dead-lorcana-import entry (P3 polish; the
Lorcana import route was gated defensively in PR #20 despite
zero current frontend callers — pages/admin/card-import.js's
<select> only offers mtg + pokemon — so if Lorcana stays
permanently out of the admin UI, this is the cleanup PR).
Added three "flagged but kept out of scope" follow-ups per the
convoy's § What did NOT change: harden-multipart-parser (P2;
5MB body still consumed before the 429 path on avatar.js),
god-function-split / refactor-cards-search-sql (P2; 240-line
7-branch SQL in cards/search.js), and withAdmin(handler) wrapper
extraction (P3 DX; the three import routes are call sites #3-5
in the codebase but uniform inline shape was preserved for
convoy atomicity). Updated tighten-visual-diff-path-filter to
note PR #20 also tripped the same false-positive.
AGENTS.md:
- Gotcha #12 extended end-to-end. Was the single-class auth-only
lib + the env-var contract; is now the 5-class reality with a
full per-class table (helper / limit-window / key / routes),
the defensive THROW pattern in extractUserIdentifier, the
gate-ordering rule for per-user limiters, and the
auth → admin-role → rate-limit ordering for the three import
routes. Prominent milestone line opens the new content:
"add-rate-limiting convoy (squash 708ef45, PR #20, 2026-05-24)
closed P0 #6 — all 8 P0s now RESOLVED." Original env-var
contract paragraph (KV_REST_API_URL / KV_REST_API_TOKEN,
fail-closed-in-prod / warn-and-noop-in-dev) is preserved
verbatim above the new content.
- § 6 Testing: intentionally untouched (no test surface changed;
vitest 21/21 and smoke 3/3 still apply).
- § 7 Deployment: intentionally untouched (no deployment-shape
changed; same KV_REST_API_* env vars from Brief 4).
.cursor/rules/api-routes.mdc:
- The implementer extended § Rate limiting in PR #20 with the
per-class table + verbatim call shape + gate-ordering rules +
identifier-extraction + uniform 429 + fail-closed env-var
contract + fail-open Upstash-outage behavior. Doc-writer pass
verified completeness; added a one-sentence convoy-attribution
line at the top of § Rate limiting citing the two-convoy
lineage (fix-auth-bypass Brief 4 for the auth class +
add-rate-limiting for the other four classes and 7 newly-gated
routes), mirroring the post-cors-tighten § CORS attribution
shape. No other touch-ups needed.
No changes to: package.json, package-lock.json, lib/rate-limit.js,
pages/**, components/**, scripts/**, test/**, tests/**,
.github/workflows/**, README.md, TESTING_GUIDE.md, playwright.config.js,
eslint.config.mjs.
Co-authored-by: Cursor <cursoragent@cursor.com>
Closes P0 #6 from PARTIAL to RESOLVED. 8/8 P0s now closed. Extends lib/rate-limit.js from single-class to 5 named limiters (auth/search/upload/generate/import). Atomically gates the 3 import routes (auth + admin-role check + rate limit) and fixes pages/admin/card-import.js's missing Bearer header in the same commit (architect's critical discovery: API gating alone would have broken the admin UI). Per Decision 1 Option A. 10 files +185/-23. Local: lint 128 baseline, vitest 21/21. CI: Playwright smoke 3/3 in 3.8s, forbidden-cors-headers pass, all gates green. PR #20 architect-commit 60b842e, implementer-commit 51a3a97. Brief 4's login.js + register.js byte-identical.
Reflects the merged cors-tighten convoy (PR #19, squash commit da50d78)
in repo documentation. Closes P0 #5 (Wildcard CORS on API surface) from
PARTIAL -> RESOLVED, leaving only P0 #6 (full add-rate-limiting) open
of the original P0 ship-blocker set. One brief in the convoy: Brief 1
shipped as planned with no scope expansions and no implementer deviations
from the verbatim spec.
.convoys/cors-tighten.md:
- frontmatter status: in-progress -> shipped (added shipped: 2026-05-24)
- new ## As-shipped section: all 5 architect-self-ratifiable decisions
ratified verbatim (D1 Option B / D2 delete OPTIONS / D3 moot / D4
no new tests / D5 add CI lock); Pattern split (16 Pattern A + 8
Pattern B) per architect's 10-file audit + implementer's per-file
diff review; diff size (25 files, +29/-261); empirical CI metrics
from post-merge run 26378806555 (forbidden-cors-headers 4s PASS,
Playwright smoke 56s 3/3 in 3.3s, Screenshot diff continue-on-error
0 with the documented Decision-4 missing-baseline failure beneath);
cross-validation that Playwright smoke continues to pass post-CORS
removal (the auth + public surfaces don't depend on the wildcard
header); implementer subagent-retry footnote (HEAD already at
a843736 when retry woke up - transient retry, work is canonical);
operator-action-required-going-forward: none; What did NOT change
audit trail.
.convoys/ship-readiness.md:
- new ## Status summary at the top (right after the code-graph line):
P0 set is now 7/8 RESOLVED; only #6 (rate-limiting) remains. Table
lists each P0 with its resolving convoy + squash commit for a quick
scan of remaining work.
- P0 #5 marked RESOLVED 2026-05-24. Added the cors-tighten as-shipped
block (24 files swept, new CI job, 16/8 Pattern split, 5 decisions
ratified, diff stat, post-merge CI metrics, transient retry
footnote, operator-action: none). Brief 4's 2026-05-23 partial
is preserved as the prior as-shipped layer above the cors-tighten
layer to maintain the audit trail.
- Queued convoys: removed the cors-tighten entry (no longer queued).
Added a new tighten-visual-diff-path-filter entry (P3 polish) -
Screenshot diff workflow triggered on API-only PR #19 because its
paths: filter is pages/** which matches pages/api/** too. ~55s of
CI waste per API-only PR; one-line YAML tweak; verify GitHub
Actions' negated-glob semantics before merging.
.cursor/rules/api-routes.mdc:
- new ## CORS section near the existing ## Dev/test endpoints
(removed) section. Documents the no-CORS-by-default convention,
the brief-4 + cors-tighten lineage, the new forbidden-cors-headers
CI gate, and three forward-conventions (no setHeader for CORS,
no OPTIONS preflight handlers, design a proper middleware layer
if a future cross-origin caller is needed - not wildcards in
individual handlers).
AGENTS.md intentionally untouched. Gotcha #5 (the public
setup-database.js endpoint) is already RESOLVED by fix-auth-bypass
Brief 3 and unrelated to this convoy. The new convention belongs in
.cursor/rules/api-routes.mdc (where API conventions live) rather than
AGENTS.md; the convoy file + the new CI gate are sufficient
documentation for the audit trail. Per convoy spec, no new gotcha
entry needed.
No changes to: package.json, package-lock.json, pages/api/**, lib/**,
components/**, scripts/**, test/**, tests/**, .github/workflows/**,
README.md, TESTING_GUIDE.md, playwright.config.js.
Co-authored-by: Cursor <cursoragent@cursor.com>
Closes P0 #5 from PARTIAL to RESOLVED. Sweeps the remaining 24 pages/api/** handlers that carried the identical scaffolded wildcard-CORS + OPTIONS preflight pattern (Brief 4 cleaned login + register; this finishes the job). Adds a blocking forbidden-cors-headers CI job modeled on forbidden-endpoints to lock the cleanup against future regression. 25 files changed (+29/-261). Local: lint 128 baseline, vitest 21/21, zero CORS matches, YAML valid. CI: Playwright smoke 3/3 in 3.3s against post-removal preview (login/verify flow still works), new forbidden-cors-headers job passes in 4s, all gates green. PR #19 architect-commit ec22b70, implementer-commit a843736.
Reflects the merged adopt-playwright-smoke convoy (PR #18, squash
commit 7b6f751) in repo documentation. Closes the test-infrastructure
side of P1 #10 step 2 (launch sequence step 10). One commit in the
convoy: Brief 1, with two small lint-baseline-preserving deviations
from the brief's verbatim shape that the implementer report flagged.
.convoys/adopt-playwright-smoke.md:
- frontmatter status: in-progress -> shipped (added shipped:
2026-05-24)
- new ## As-shipped section: operator-ratified Decisions (D1 keep
.ts, D4 defer baselines, D6 simple scripts); two implementer
deviations (removed unused eslint-disable-no-console directive
that would have regressed lint 128 -> 129; placed @playwright/test
first in devDeps for alphabetical correctness); cross-validation
that smoke test 2 ("sign-in page renders") locks in PR #15's
logged-out CTA work in components/Layout.js; empirical metrics
from post-merge run 26376162598 (59s workflow, 3/3 in 2.9s,
0-leak); operator-action-required note pointing at the queued
seed-visual-baselines-on-linux follow-up; What did NOT change
audit trail.
.convoys/ship-readiness.md:
- Queued convoys: new entry seed-visual-baselines-on-linux (Linux-
Docker baseline generation per Decision 4 + Boot-the-brief
Finding 7; Mac-generated baselines would silently overwrite Linux
CI baselines because the custom snapshotPathTemplate has no
{platform} token).
- Queued convoys: new RESOLVED block for adopt-playwright-smoke
(PR #18, 7b6f751) — as-shipped surface, implementer deviations,
empirical metrics (59s workflow / 3/3 in 2.9s / 0 secret leaks),
the PR #15 cross-validation finding, operator-action-required
going forward (the seed-visual-baselines-on-linux follow-up),
flagged-but-deferred items, and ownership trail (3 architect-
self-ratified decisions + 3 operator-ratified).
- Launch sequence step 10: marked RESOLVED 2026-05-24 with the
commit + metrics inline.
- P1 #10 No tests Fix sequence: step 2 marked RESOLVED with the
convoy + metrics ref; step 3 (re-enable test: job in
ci.yml) called out as the next remaining task; step 5 (wire
preview-smoke.yml to the Vercel preview URL) marked RESOLVED
across PR #17 + PR #18 since both contributed.
AGENTS.md:
- Section 6 Testing: rewritten end-to-end. Was "E2E/smoke runner
still pending"; is now "@playwright/test@^1.60.0 wired, two
projects (smoke + visual), npm run test:smoke / test:visual /
test:visual:update". Documents the local-run convention (boot
next dev separately, then BASE_URL=... npm run test:smoke);
the one-time npx playwright install --with-deps chromium step;
the no-baselines-yet state + the Linux-Docker seed command +
the cross-platform mismatch reason (no {platform} token in
snapshotPathTemplate); the CI behavior split (vitest blocking,
smoke on every PR with pipeline:skip-smoke escape hatch,
Screenshot diff path-filtered with the first-red-on-missing-
baseline state documented). Updates vitest coverage count
16 -> 21 (the +5 Layout regression-lock tests from PR #15).
Notes TESTING_GUIDE.md is being eclipsed and will be renamed
to docs/MANUAL_QA.md in a future cleanup convoy.
- Section 7 Deployment: rewrites the Vercel-bypass paragraph from
a single "query param now / header reserved for future" bullet
into a two-shape audit ((1) query param on the wait-action's
path: input per PR #17; (2) HTTP header in playwright.config.js's
use.extraHTTPHeaders per PR #18). Documents the Decision-2
fail-loud-in-CI / warn-in-dev predicate and references Gotcha #12
as the established precedent (lib/rate-limit.js). Picked Section
7 over a new Gotcha because the bypass plumbing is operationally
a deployment concern, not an app pitfall.
No changes to: package.json, package-lock.json, playwright.config.js,
eslint.config.mjs, lib/**, pages/**, components/**, scripts/**,
.github/workflows/**, .cursor/rules/**, README.md,
tests/visual/homepage.spec.ts (JSDoc is already neutral-tense, no
future-tense references to clean up).
Co-authored-by: Cursor <cursoragent@cursor.com>
Closes P1 #10 step 2. Adds @playwright/test@^1.60.0, playwright.config.js (CI fail-loud / dev warn-and-continue for missing VERCEL_AUTOMATION_BYPASS_SECRET, two projects partitioned by testMatch, snapshotPathTemplate aligned with workflow upload path), tests/visual/homepage.spec.ts (1 test, no baseline committed per architect Boot-the-brief Finding 7), 3 npm scripts, and 3 .gitignore entries. Smoke tests now run end-to-end against Vercel preview with x-vercel-protection-bypass header: 3/3 passed in 2.9s, total workflow 59s. Zero secret leaks in log. PR #18 architect-commit 3ac527e, implementer-commit c72d006.
Reflects the merged fix-vercel-deployment-protection-in-ci convoy
(PR #17, squash commit 9a3e077) in repo documentation. Closes the
CI-infra side-effect of P0 #7. Three commits in the convoy: Brief 1
(bypass plumbing per spec), plus two scope expansions discovered
during CI validation (shell-injection hardening, and a Decision-A
shape correction to drop the cookie-bypass param).
.convoys/ship-readiness.md:
- Queued convoys: mark fix-vercel-deployment-protection-in-ci as
RESOLVED 2026-05-24 with 9a3e077. Document the 3-commit reality
(365e9f0 Brief 1 bypass plumbing, b6f8688 shell-injection
hardening of the gate Decide step, 043a6ee dropping
&x-vercel-set-bypass-cookie=true), the empirical metrics
(wait-action: 10-min timeout -> 194ms; workflow runtime:
10+ min -> 59s), the documented expected red on Playwright
smoke (npx playwright test fails because playwright.config.js
doesn't exist yet -- adopt-playwright-smoke owns that), and the
operator-rotation caveat (R6).
AGENTS.md:
- Section 7 Deployment: correct the noun "header" -> "query param
on wait-action's path:" since that's what actually landed per
Decision A. Also document the without-cookie form (the cookie
variant 401s through axios's missing cookie jar) and the
operator re-seed runbook for token rotation.
- Section 7 Deployment: fold in a one-liner about the GitHub
Actions ${{ }}-in-shell-is-injection-vector pattern, with the
env: + quoted-shell fix shape. Picked Section 7 over a new
Gotcha #13 because the existing Gotchas list is dominated by
app-level pitfalls (auth, SQL clients, ESLint), and CI YAML
hardening is naturally co-located with deployment.
.convoys/fix-vercel-deployment-protection-in-ci.md:
- frontmatter status: in-progress -> shipped (added shipped:
2026-05-24)
- new ## As-shipped section: 3-commit reality, Decision-A shape
deviation (we shipped without &x-vercel-set-bypass-cookie=true),
empirical timings (194ms wait, 59s total), remaining-red
attribution to adopt-playwright-smoke, and the
operator-rotation caveat.
No changes to: package.json, lib/**, pages/**, components/**,
scripts/**, .github/workflows/**, .cursor/rules/**, README.md.
Co-authored-by: Cursor <cursoragent@cursor.com>
* architect: plan fix-vercel-deployment-protection-in-ci convoy (1 brief)
Flip convoy status queued -> in-progress and append the architect
output per the role-architect contract:
- Decisions (post-IA round) A/B/C/D with recommendations + rationale,
routed back to operator for gate-1 ratification (A and D especially).
- Architecture section: file plan, risk list (R1-R8), test plan,
decomposition (1 brief, justified), slice_dependencies block.
- brief-1: inject VERCEL_AUTOMATION_BYPASS_SECRET into both workflows
via query-param-on-path (Decision A), tighten max_timeout 600 -> 120
(Decision B), extend gate to skip fork PRs (Decision D), and forward
the secret as an env var to the Playwright/visual step for forward-
compat with adopt-playwright-smoke.
Boot-the-brief findings preempted in the brief:
- wait-for-vercel-preview@v1.3.2 source confirms only `targetUrl` (the
bare deployment URL) is ever logged or emitted as outputs.url; the
bypass query in `path:` is structurally invisible to logs and to
downstream consumers of ${{ steps.vercel.outputs.url }}.
- The action exposes no custom-header input -- Option B is mechanically
impossible for the wait step without forking the action.
- `path:` MUST begin with a leading `/` because the action parses it
via `new URL(path, url)`.
- PR #16 run logs confirm the 401 timeout failure mode the convoy targets.
- Playwright config doesn't exist yet (adopt-playwright-smoke owns it);
this brief only plumbs the env var.
No workflow YAML, no Playwright config, no AGENTS.md changes in this
commit -- those land in the implementer phase per the architect contract.
Awaiting human gate 1 (Decisions A + B + D ratification + brief approval)
before implementer dispatch.
Co-authored-by: Cursor <cursoragent@cursor.com>
* fix(ci): plumb VERCEL_AUTOMATION_BYPASS_SECRET into preview-smoke + visual-diff (Brief 1 of fix-vercel-deployment-protection-in-ci)
Closes the CI-infra half of P0 #7's follow-up. PR #16 (squash commit
7e97254) added scoped permissions to both workflows but exposed that
Vercel Deployment Protection 401s anonymous GitHub-runner requests,
causing both Playwright smoke and Screenshot diff to time out at 10
minutes on every PR. This brief plumbs the bypass secret end-to-end
so the wait-action's healthcheck reaches 200.
Per architect Decision A (.convoys/fix-vercel-deployment-protection-in-ci.md):
- preview-smoke.yml + visual-diff.yml: wait-for-vercel-preview's
`path:` input now carries the bypass as a query parameter
`?x-vercel-protection-bypass=${{ secrets.* }}&x-vercel-set-bypass-cookie=true`.
The action only logs the bare targetUrl (verified in action.js:357,360,363)
so the secret stays out of workflow logs.
Per Decision B:
- max_timeout: 600 -> 120. PR #16 evidence shows Vercel previews are up
within seconds of job start; 120s gives ample headroom and surfaces
misconfigurations in ~2 minutes instead of ~10.
Per Decision D (NEW -- surfaced by Boot-the-brief):
- gate: job's Decide step now checks github.event.pull_request.head.repo.fork
FIRST. Forks lack repo secrets, so they would otherwise burn ~4 minutes
per PR on a misleading 401. The fork-check emits `::notice::` and short-
circuits before the existing skip-via-PR-body directive runs.
Forward-compat for adopt-playwright-smoke:
- Both workflows' Playwright/screenshot-capture step now exports
VERCEL_AUTOMATION_BYPASS_SECRET as env. The actual Playwright config
consumes it via extraHTTPHeaders in adopt-playwright-smoke's brief.
BASE_URL stays as the bare \${{ steps.vercel.outputs.url }} (no query
string) so it remains safe to echo.
Verification:
- YAML parses (js-yaml load on both files: preview-smoke jobs [gate, smoke];
visual-diff jobs [gate, visual])
- actionlint not run (binary not installed locally); recommend installing
for future PRs. Future adopt-actionlint convoy could add it to CI.
- No `set -x`, `echo`, `cat`, or `printf` of the secret or bypass URL
in any modified step
- permissions: and concurrency: blocks unchanged (PR #16 contracts preserved)
Co-authored-by: Cursor <cursoragent@cursor.com>
* fix(ci): route github.event.pull_request.body through env: to avoid shell injection
Pre-existing latent bug surfaced by PR #17's CI run. The Decide step's
inline `${{ github.event.pull_request.body }}` interpolation pastes
arbitrary PR-body text directly into a bash script. When the body
contains shell metacharacters (parens, pipes, backticks, redirections,
etc.) the resulting script either errors out at YAML-load time OR —
worse — executes attacker-controlled shell.
This bit PR #17 with a real syntax error ("unexpected token `('")
because the PR body contains parenthesized phrases like
"(was: 10-minute timeout)". Every Decide-step run in this repo has
been one badly-formatted PR body away from breaking the gate.
Fix: forward `github.event.pull_request.body` and
`github.event.pull_request.head.repo.fork` through the step's `env:`
block as `PR_BODY` and `PR_IS_FORK`, then quote them in shell
(`"$PR_BODY"`, `"$PR_IS_FORK"`). The env-var path leaves the values
as plain strings rather than syntactically embedded code, which is the
standard GitHub Actions hardening pattern (see GitHub's "Security
hardening for GitHub Actions" → "Using a third-party action").
Same change in both workflows; ~9 LOC each.
This fix is technically beyond Brief 1's scope (which targeted only
Vercel-bypass plumbing) but is added in this convoy because the bug
actively blocks Brief 1's success criterion from being validated on
PR #17. Documented in the convoy file's "Anything flagged but not
acted on" follow-up pass.
Co-authored-by: Cursor <cursoragent@cursor.com>
* fix(ci): drop x-vercel-set-bypass-cookie=true from wait-action path
The wait-action's healthcheck was still 401ing despite the bypass query
being correct. Root cause: with `x-vercel-set-bypass-cookie=true`, Vercel
returns 307 + Set-Cookie (`_vercel_jwt`), but axios in Node has no cookie
jar — it follows the redirect to the bare URL without the cookie, which
then 401s.
Local verification (run by operator):
curl -sI "https://<preview>/?x-vercel-protection-bypass=<secret>" | head -1
→ HTTP/2 200 (works, no cookie needed)
curl -sI "https://<preview>/?x-vercel-protection-bypass=<secret>&x-vercel-set-bypass-cookie=true" | head -1
→ HTTP/2 307 (the redirect-without-cookie path that breaks axios)
For a one-shot healthcheck, the per-request bypass query is enough.
The cookie variant stays reserved for the future Playwright config
(adopt-playwright-smoke) where a real browser cookie jar exists.
Added an inline comment in preview-smoke.yml explaining this so the
next agent doesn't accidentally re-add the cookie param.
Co-authored-by: Cursor <cursoragent@cursor.com>
---------
Co-authored-by: Cursor <cursoragent@cursor.com>
Doc-drift fix discovered after the fix-layout-default-user cleanup
pass. The bump-next-js convoy shipped on 2026-05-23 as commit
e57ea17, and the commit body explicitly states "Closes P0
ship-blocker #8" — but no dedicated doc-writer pass ever ran for
that convoy, so .convoys/ship-readiness.md still had P0 #8 listed
as open even though Vercel has been deploying main + every PR
successfully since.
Patches only P0 #8 to RESOLVED with the same shape used by P0
#1, #2, #3, #4, #7:
- Title: "— **RESOLVED 2026-05-23**"
- Resolved by: bullet citing e57ea17
- As-shipped: 7-point summary of the bump's actual deliverables
(next 15.4.3->16.2.6, ESLint v9-fallback, typescript devDep,
eslint.config.mjs, scripts.lint update, images.remotePatterns
migration, build verification)
- Side-effects deferred: bump-react, bump-eslint-10, bump-typescript-6,
fix-lint-baseline, App Router migration
- Doc drift note acknowledging this entry was added ~24h late
No code changes. No package.json / lib/** / pages/** / components/**
touched. AGENTS.md was already current (Gotchas #9, #10, #11 already
reflect Next 16 + ESLint v9 + Turbopack post-bump state).
Co-authored-by: Cursor <cursoragent@cursor.com>
Reflects the merged fix-layout-default-user convoy (PR #15) and the
companion CI permissions fix (PR #16) in repo documentation. Also
queues the new fix-vercel-deployment-protection-in-ci convoy that
PR #16 exposed.
.convoys/ship-readiness.md:
- P0 #7: mark RESOLVED 2026-05-24 with squash commit ca302a8.
Document the as-shipped Layout default-null change, the 7-page
sweep, the 5 new regression-lock vitest assertions, and the
queued follow-ups (single-auth-provider, MobileNavigation
cleanup) that stayed explicitly out of scope.
- Queued convoys: add fix-vercel-deployment-protection-in-ci
(P2, CI infra) — PR #16's permissions fix exposed that Vercel
Deployment Protection 401s anonymous CI requests; needs a
bypass-secret plumb to land cleanly. New section also captures
other in-flight follow-ups (rotate-default-admin, cors-tighten,
add-rate-limiting, purge-weak-creds-from-helpers,
single-auth-provider, cleanup-mobile-nav-dead-props,
bump-eslint-10) so the audit trail is centralized.
AGENTS.md:
- § 4 Gotcha #8: mark RESOLVED with commit ref ca302a8. Mirror
the convention used by the prior cleanup commits for #2, #3,
#4, #5 (entry kept, not renumbered).
- § 3 Key conventions: add a new "Layout user prop" bullet
documenting the new default-null + logged-out-CTA contract so
the convention is discoverable from the conventions list, not
just the resolved-gotcha entry.
.convoys/fix-vercel-deployment-protection-in-ci.md (new):
- Queued scaffold. Operator must seed
VERCEL_AUTOMATION_BYPASS_SECRET as a repo secret before the
implementer can run. Decisions to ratify (query param vs.
header), known constraints, acceptance criteria, and
out-of-scope all enumerated.
No changes to: package.json, lib/**, pages/**, components/**,
scripts/**, .github/**, README.md, .cursor/rules/**.
Co-authored-by: Cursor <cursoragent@cursor.com>
* convoy: scope fix-layout-default-user (P0 #7 — Layout maintainer-email leak)
The last remaining P0 ship-blocker from .convoys/ship-readiness.md.
components/Layout.js line 562 defaults the user prop to a real email
address (me@randallstillwell.com); any page that renders Layout without
passing user explicitly impersonates the maintainer.
Scope: components/Layout.js + audit of 17 pages that import Layout
(grep-confirmed list in convoy file). Single PR likely. Auditor cohort
skipped (no design-system, IA, or browser-smoke surface).
Architect to address:
- Q1: logged-out rendering branch design (navbar, mobile-nav,
auth-only items treatment)
- Q2: page audit triage into always-auth / public-or-auth /
anonymous-allowed buckets
- Q3: brief decomposition (single brief / 2 briefs in 1 PR / fan-out)
- Q4: whether to add vitest coverage for the logged-out branch
(recommend yes — small surface, high regression protection)
Hard out-of-scope: branding (pick-a-name), auth-provider collapse
(single-auth-provider), Layout god-component split (god-component-split).
depends_on: bump-next-js (shipped), fix-auth-bypass (shipped),
drop-public-setup (shipped)
addresses: P0 #7 from .convoys/ship-readiness.md
parent: ship-readiness
Co-authored-by: Cursor <cursoragent@cursor.com>
* architect(fix-layout-default-user): plan + briefs 1-2 (Layout fix + page audit)
2 briefs, single PR. ~12 files net (down from the 18 in the original scope —
10 of the 17 Layout-importing pages already pass user explicitly).
Brief 1: components/Layout.js default user=null + Sign-in CTA branch in
UserProfileDropdown when logged out. Adds first jsdom test in the repo
at test/components/Layout.test.js (Decision D2) with 5 regression-lock
assertions. devDeps: jsdom@^29, @testing-library/react@^16.
Brief 2: page audit sweep — 7 pages need code changes:
- Pass user={user} to Layout: scanner.js, deck-builder.js (×4),
deck/[id].js (×3), decks.js (×3)
- Replace page-level useState({email: 'me@...'}) → useState(null) +
null-guards: profile.js, settings.js
- Replace hardcoded const user = {email: 'me@...'} with useAuth():
card/[id].js
Discovered second anti-pattern: profile.js, settings.js, card/[id].js
seed page-level state with the maintainer email. Folded into Brief 2 since
success metric "no real email address remains in any component default-prop"
reads naturally to include page-level seed values.
Decisions:
A1 — Sign-in CTA replaces avatar+email+dropdown when user===null;
hides auth-only dropdown (Profile/Settings/Logout/Admin);
keeps public + community nav visible
B — Per-page bucket assignment (10 already correct, 7 need fix);
full per-page table with justification in convoy file
C2 — Two briefs in one PR (Brief 1 = Layout + test; Brief 2 = page
sweep depends on Brief 1). C1 buries the conceptual change under
mechanical edits; C3 is over-orchestrated for this scope
D2 — vitest lock-in; first jsdom test in repo; same negative-regression
style as test/lib/permission-middleware.test.js (synthetic-admin
shape). devDeps jsdom + @testing-library/react
Risks tracked R1-R8. Biggest: R2 (useState(null) null-deref in 3 leaky
pages — mitigated by audit-pass mandate + manual smoke).
MobileNavigation deliberately NOT folded in: its user prop is dead code
(never reads user.*); different bug class; cleanup queued separately to
avoid scope expansion.
Flagged-but-deferred:
- 4 pages still import useAuth from lib/auth-context.js
→ single-auth-provider (queued P1 #9)
- Layout headers still render "Deck Hearth" / "DH" branding
→ pick-a-name (queued P1 #12)
- MobileNavigation dead user prop → cleanup-mobile-nav-dead-props
or fold into god-component-split
addresses: P0 #7 from .convoys/ship-readiness.md (last P0 ship-blocker)
parent: ship-readiness
Co-authored-by: Cursor <cursoragent@cursor.com>
* feat(layout): default user=null + Sign-in CTA when logged out (Brief 1 of fix-layout-default-user)
Closes the source-side half of P0 #7 from .convoys/ship-readiness.md.
The page-side sweep (Brief 2) follows in a separate commit.
components/Layout.js:
- Default user prop is now null (was hardcoded to
{ email: 'me@randallstillwell.com', role: 'user' })
- UserProfileDropdown renders a "Sign in" link to /login when
user === null instead of the maintainer's email + auth-only menu
items (Decision A1)
- All user.* accesses guarded with optional chaining or null checks
- useState hook stays above the new null-user early return to satisfy
rules-of-hooks (boot-the-brief caught this on the first try;
see AGENTS.md Gotcha #11.5)
test/components/Layout.test.js (new):
- First jsdom test in the repo (Decision D2)
- 5 regression-lock assertions: no maintainer email ever rendered
(prop omitted, prop=null), Sign-in link exists with href=/login,
supplied email renders when prop is set, no "Guest" placeholder
(locks A1 copy choice)
- Mocks next/link, next/router (prefetch, replace, events, query),
and theme-context.useTheme for jsdom safety under Next 16
package.json + package-lock.json:
- Add jsdom@^29 and @testing-library/react@^16 to devDependencies
- @testing-library/dom@^10 added explicitly (peer auto-install
skipped it under npm 11; brief anticipated this fallback)
vitest.config.js (deviation from brief — see PR description):
- Add esbuild { loader: 'jsx', jsx: 'automatic' } so vitest can
parse JSX in .js files. Required to import any React component
written in the repo's Next.js pages-router .js convention
(AGENTS.md Gotcha #9). The brief said "no change" to this file,
but JSX-in-.js parsing is a hard prerequisite for the new test
to import components/Layout.js — the alternatives (rename test
to .test.jsx; rewrite test in React.createElement) either break
the test glob or still hit the same Layout.js parse failure.
Other tests are unaffected (they import non-JSX modules).
Smoke output: see PR description.
addresses: P0 #7 from .convoys/ship-readiness.md (last P0 ship-blocker)
Co-authored-by: Cursor <cursoragent@cursor.com>
* feat(pages): pass user explicitly + null-guard leaky page seeds (Brief 2 of fix-layout-default-user)
Closes the page-side half of P0 #7 from .convoys/ship-readiness.md.
Brief 1 (commit ddf8fd2) handled the Layout-side fix.
Per the architect's per-page bucket table (Decision B in
.convoys/fix-layout-default-user.md), 7 pages needed code changes;
the other 10 of 17 Layout-importing pages already pass `user` correctly.
Pass user={user} to Layout (4 pages, 11 call sites):
- pages/scanner.js (1 call)
- pages/decks.js (3 calls)
- pages/deck-builder.js (4 calls)
- pages/deck/[id].js (3 calls)
(All four still import useAuth from lib/auth-context.js — that's
intentional and stays as-is until the single-auth-provider convoy
collapses the three parallel auth surfaces.)
Replace leaky page-level seed values with useState(null) + null guards
(2 pages, R2 mitigation):
- pages/profile.js: useState({email: 'me@...', role: 'user', ...})
→ useState(null) + ?. on every sync user.* read
+ early-return guards in getDisplayName/getInitials
+ conditional render around the "Member since" block
so formatDate(undefined) never runs
- pages/settings.js: same pattern (single user.email reader guarded)
Replace hardcoded const with useAuth from lib/use-auth.js (1 page):
- pages/card/[id].js: const user = {email: 'me@...'}
→ const { user } = useAuth() (called unconditionally
at the top of the component; rules-of-hooks safe)
Verification:
- grep 'me@randallstillwell.com' pages/ → 0 hits
- 21/21 vitest tests pass (16 pre-existing + 5 from Brief 1)
- npm run lint matches baseline (128 problems pre, 128 post; verified
via git stash before/after)
- Manual static read-through of every diff; ReadLints clean on the 7
files
- Dev-server smoke: /cards anonymous returned HTTP 200 with 0
'me@randallstillwell' matches before the user's shared dev server
became unresponsive mid-session (same dev-server-shared-by-user
constraint flagged in Brief 1); interactive logged-in smoke is
parent/operator gated
Flagged-but-deferred (untouched per scope):
- 4 pages still import useAuth from lib/auth-context.js
→ single-auth-provider (queued P1 #9)
- components/MobileNavigation.js still receives dead user prop
→ cleanup-mobile-nav-dead-props (or fold into god-component-split)
addresses: P0 #7 from .convoys/ship-readiness.md (last P0 ship-blocker)
Co-authored-by: Cursor <cursoragent@cursor.com>
---------
Co-authored-by: Cursor <cursoragent@cursor.com>
Both Playwright-on-Vercel workflows fail at the very first action
(`patrickedqvist/wait-for-vercel-preview@v1.3.2`) with 403 "Resource
not accessible by integration". The cause is the repo-default workflow
token being read-only-by-default with no scopes declared by the workflow.
Add minimal scoped permission blocks per the GitHub Actions least-privilege
guidance:
preview-smoke.yml:
contents: read
deployments: read # wait-for-vercel-preview queries GitHub Deployments
pull-requests: read # correlate deployment with this PR
statuses: read # some Vercel deployments use commit statuses
visual-diff.yml:
contents: read
deployments: read
pull-requests: write # final step posts "Visual Diff" comment via Issues API
statuses: read
Verified against the failure on PR #15:
- Playwright smoke: "Resource not accessible by integration" on
wait-for-vercel-preview → fixed by deployments+statuses+pull-requests:read
- Screenshot diff: 403 from POST /repos/.../issues/15/comments with
`x-accepted-github-permissions: issues=write; pull_requests=write`
in the response → fixed by pull-requests:write (covers Issues API
for PR comments; issues:write would also work but pull-requests:write
is the idiomatic scope)
Unblocks visual-regression signal on every future PR. No code changes,
no test changes — workflow YAML only.
Co-authored-by: Cursor <cursoragent@cursor.com>
Reflects the merged drop-public-setup convoy (PR #13) in repo
documentation. Small focused pass; no behavior changes.
AGENTS.md:
- § 1 auth bullet: replace "seed admin row still ships in
setup-neon-db.js" claim with the new env-var-gated reality and
the R1 operator-rotation caveat.
- § 4 Gotcha #4: mark RESOLVED with commit refs (ff80753 + b63b509),
document the as-shipped behavior, the Brief 2 CJS→ESM Node 22.x
fix, and the R1 operator caveat. Entry kept (not renumbered) per
the same convention used for resolved gotchas #2, #3, #5.
- § 5 Running locally: add ADMIN_INITIAL_PASSWORD to the env-var
template list with a note that setup-db exits 1 if it's unset.
.convoys/ship-readiness.md:
- P0 #3: mark RESOLVED 2026-05-23 with commit refs, document the
full as-shipped behavior including Brief 2's CJS→ESM bonus,
the R1 operator caveat (Decision A — going-forward only),
and the deferred sibling weak-cred references queued for
purge-weak-creds-from-helpers.
.cursor/rules/no-go-zones.mdc:
- Editing rules of thumb: clarify the schema-vs-operational
distinction for scripts/setup-neon-db.js. drop-public-setup
set the precedent that operational changes (env-var gating,
pre-flight validation, module-system fixes) are allowed in
place, while DDL changes still need a separate migration
script. Documented so future agents don't have the same
confusion the drop-public-setup architect did (see Decision B
in .convoys/drop-public-setup.md).
No changes to: package.json, lib/**, pages/**, components/**,
scripts/**, .github/**, README.md (already updated in PR #13).
Co-authored-by: Cursor <cursoragent@cursor.com>
Brief 2 of drop-public-setup. Closes Decision D — pure module-system
conversion of scripts/setup-neon-db.js so `npm run setup-db` actually
runs on Node 22.x where package.json has "type": "module" (added by
bump-next-js for ESLint v9 flat-config support).
Before this commit, `npm run setup-db` threw:
ReferenceError: require is not defined in ES module scope
After this commit, brief 1's ADMIN_INITIAL_PASSWORD env-var gate actually
fires as documented.
Changes (all in scripts/setup-neon-db.js):
- require('dotenv').config(...) → import dotenv + dotenv.config(...)
- require('@neondatabase/serverless') → import { neon }
- inline require('bcryptjs') hoisted to top-of-file import bcrypt
- no functional changes; same DDL, same env-var gate, same console.logs
Smoke verification: see PR description.
Co-authored-by: Cursor <cursoragent@cursor.com>
Mid-convoy discovery: brief 1's implementer confirmed scripts/setup-neon-db.js
does not actually run on Node 22.x because bump-next-js added "type": "module"
to package.json but the seed script still uses CJS require() calls.
Throws ReferenceError immediately on `npm run setup-db`.
Architect's original "Anything flagged but not acted on" #1 claim that "it
runs successfully today under Node 22" was incorrect for Node 22.14.0.
Decision D (ratified by user 2026-05-23): expand convoy to include brief 2
rather than queue a separate convert-setup-db-to-esm follow-up. Rationale:
brief 1's env-var gate is theatrical security on a script no operator can
execute; the CJS→ESM conversion is mechanical (~6 LOC, same file, no
functional changes); splitting into two convoys creates a regression window
where operators on Node 22.x cannot bootstrap a database.
Brief 2 scope: pure module-system conversion in scripts/setup-neon-db.js:
- require('dotenv').config(...) → import dotenv + dotenv.config(...)
- require('@neondatabase/serverless') → import { neon }
- inline require('bcryptjs') hoisted to top-of-file import
- no functional changes; same DDL, same env-var gate, same console.logs
Verification: smoke must now show npm run setup-db actually executes (no
ReferenceError); brief-1 env-var gate must still fire as documented;
all 16 vitest tests must still pass.
Updated:
- .convoys/drop-public-setup.md Decomposition (brief 2 added, depends_on brief 1)
- .convoys/drop-public-setup.md slice_dependencies YAML
- .convoys/drop-public-setup.md § Decisions (added Decision D)
- .convoys/drop-public-setup.md § Anything flagged but not acted on
(item #1 marked resolved by brief 2)
addresses: P0 #3 from .convoys/ship-readiness.md + Node 22.x compat
parent: ship-readiness
Co-authored-by: Cursor <cursoragent@cursor.com>
Closes P0 #3 from .convoys/ship-readiness.md.
scripts/setup-neon-db.js:
- Read ADMIN_INITIAL_PASSWORD env var at the top of setupNeonDatabase()
before any DB connection. Fail loudly (process.exit(1)) with an
actionable message if unset or empty.
- Replace bcrypt.hash('admin123', 12) with bcrypt.hash(adminPassword, 12).
- Delete the two console.log lines that echoed admin user + password to
stdout (R3 - stdout leak into CI logs).
- Keep ON CONFLICT (email) DO NOTHING unchanged. Re-running setup-db
on an env with the admin row already present is a no-op for the
password (R4 - silent rotation prevention). Rotation of existing
weak-hash admin rows is out of scope (Decision A - queued for the
rotate-default-admin follow-up convoy).
README.md:
- Add ADMIN_INITIAL_PASSWORD to the install-step env-example block
with a CI-secret note (and add KV_REST_API_URL/KV_REST_API_TOKEN
for completeness; they're optional for local dev).
- Replace the "Default Admin Account" section with "First-time
admin setup", documenting the env var, openssl rand suggestion,
and the operator rotation note for envs that predate this change.
- Zero occurrences of 'admin123' remain in README.md (the operator
rotation note refers to "the prior weak default" instead of naming
the literal string, so grep verification A2 holds).
Decisions A1 (going-forward only), B (operational change allowed),
C1 (no vitest coverage - manual smoke in PR description) per
.convoys/drop-public-setup.md section Decisions.
Smoke output: see PR description.
Co-authored-by: Cursor <cursoragent@cursor.com>
Single-brief convoy. ~25 LOC net across 2 files (scripts/setup-neon-db.js,
README.md). No fan-out; single PR.
Decisions:
A1 — going-forward only (matches JWT_SECRET pattern from
fix-auth-bypass Brief 1). Existing weak-hash admin rows in
deployed envs are NOT rotated; operators rotate manually
via the app after merge. Queue rotate-default-admin follow-up
if a real audit finds a deploy still on the weak hash.
B — operational change to setup-neon-db.js is allowed; no-go-zones
rule prohibits SCHEMA edits, not env-var gating.
C1 — no vitest coverage. Fail-loud path is validated by manual smoke
(the brief mandates pasting fail-loud + happy-path output into
the PR description).
Key risks tracked: R1 (existing weak hash), R2 (unhelpful error),
R3 (stdout password leak — delete the console.log lines, do NOT
interpolate the env-var), R4 (silent rotation if ON CONFLICT changed
to DO UPDATE), R5 (README env block omits the new var), R6 (3 sibling
files still have admin123 — out of scope per convoy spec).
Flagged-but-deferred:
- CommonJS in ESM package (setup-neon-db.js) → convert-setup-db-to-esm
- admin123 in reset-db.js, create-test-users.js, TESTING_GUIDE.md
→ purge-weak-creds-from-helpers (or fold into launch-polish)
- admin@tcgvault.com hardcoded email → pick-a-name convoy
addresses: P0 #3 from .convoys/ship-readiness.md
parent: ship-readiness
Co-authored-by: Cursor <cursoragent@cursor.com>
Closes P0 #3 from .convoys/ship-readiness.md. Removes the hardcoded
admin@tcgvault.com / admin123 credentials from scripts/setup-neon-db.js
(make it env-var-driven) and from README.md.
Single PR; small surface (2 files). No UI, no API, no migration. Auditor
cohort skipped (no UX/a11y/design-system surface).
P0 #3's third file (pages/api/setup-database.js) was already deleted by
fix-auth-bypass Brief 3.
Architect to address:
- Existing-admin rotation story (going-forward fix vs. forced reset)
- Confirm scripts/setup-neon-db.js no-go-zones rule applies to schema
changes only (this is operational)
- Whether to add vitest coverage for the env-var-required path
depends_on: bump-next-js (shipped), fix-auth-bypass (shipped)
Co-authored-by: Cursor <cursoragent@cursor.com>
Closes AGENTS.md gotcha #11 (well, the relevant half of it — "Testing:
None yet" line in §6 is now stale).
Installs vitest@^3.2.4 (single devDep, no UI / coverage / jsdom) and
adds 16 unit tests across 3 files that lock in post-Brief-1/2/4
behavior:
test/lib/auth-secret.test.js (3 tests)
- JWT_SECRET exports the env value
- JWT_TOKEN_TTL is canonical 24h
- Module throws at load when JWT_SECRET is empty
test/lib/permission-middleware.test.js (8 tests)
- getUserFromRequest returns null for: missing header, non-Bearer
scheme, malformed token, wrong-secret token, expired token,
valid-token-no-user-row
- Returns user object for valid token + user row
- Brief 2 regression lock: does NOT return the synthetic admin
shape { userId: 1, email: 'admin@tcgvault.com', role: 'admin' }
when no Authorization header is present
test/api/auth-utils.test.js (5 tests)
- generateToken issues 24h JWT (exp - iat === 86400)
- Payload includes userId, email, role
- verifyToken round-trips valid tokens
- Returns null for malformed / wrong-secret tokens
CI: re-enabled the previously commented-out test: job in
.github/workflows/ci.yml. Blocking (no || true wrapper) — vitest is
the first runner in this repo and we want CI red on test regression.
JWT_SECRET is set via a CI-only fake; production secret is unaffected.
Rate-limit (Brief 4) coverage deferred to a future expand-auth-tests
convoy per architect's call (R11). package.json has "type": "module"
so vitest's default Vite-based transform handles .js ESM out of the
box — no transform config needed.
Convoy: fix-auth-bypass / Brief 5 (last brief)
Co-authored-by: Cursor <cursoragent@cursor.com>
Follow-up to fix-auth-bypass Brief 2 (commit 258e479). Brief 2 made
getUserFromRequest return null for unauthenticated requests. POST, PUT,
and DELETE branches of pages/api/collections/[identifier]/cards.js
were dereferencing user.userId without a guard → NPE → HTTP 500.
Security side was already fixed by Brief 2 (no more
anonymous-write-as-admin on collections owned by userId: 1). This patch
adds the cosmetic 500 → 401 cleanup the Brief 2 reviewer flagged.
Three identical 'if (!user) return 401' guards added, one per write
branch. GET branch was already guarded via the ternary pattern.
Sibling endpoints under pages/api/collections/** were re-audited by the
implementer and confirmed correctly guarded (thumbnails, permissions,
activity all have early null checks; [identifier].js uses optional
chaining throughout). No further hotfixes needed for that route group.
Convoy: fix-auth-bypass / Brief 6 (post-architect hotfix)
Co-authored-by: Cursor <cursoragent@cursor.com>
Adds rate limiting to /api/auth/login and /api/auth/register and removes
their wide-open CORS allowlist.
Rate limiting (@upstash/ratelimit + @upstash/redis):
- 5 attempts per 15-minute sliding window per IP, prefix "tcgvault:auth"
- new lib/rate-limit.js, lazy singleton, single source of truth
- reads KV_REST_API_URL / KV_REST_API_TOKEN (Vercel Upstash Marketplace
convention — auto-provisioned, no manual env-var setup needed)
- fail-closed in production if env vars are missing (better to error
one login than silently disable brute-force protection on live)
- fail-open in dev/test if env vars are missing (single console.warn)
- fail-open on Upstash backend outage (defense-in-depth — don't lock
the entire userbase out if Upstash is down)
- IP extracted from x-forwarded-for first hop, with socket fallback;
NOT req.body.email (rotates) or Authorization header (absent on
unauthenticated login)
CORS:
- Removed Access-Control-Allow-Origin: * + companion headers + OPTIONS
preflight from login.js and register.js
- These are first-party endpoints called from the same-origin SPA; the
"*" allowlist was a development convenience that shipped to prod
- verify.js is OUT OF SCOPE per architect's "cors-tighten" deferral
(see convoy plan § Architect's calls)
Other handler ordering preserved verbatim per brief: method gate first,
then rate-limit check (returns 429 with Retry-After header), then the
existing try/catch + body parsing + DB work.
Pre-merge requirements: KV_REST_API_URL + KV_REST_API_TOKEN must be set
in Vercel Production (already done — Upstash marketplace integration
auto-provisioned both, confirmed by maintainer 2026-05-23).
Convoy: fix-auth-bypass / Brief 4
Co-authored-by: Cursor <cursoragent@cursor.com>
Closes AGENTS.md gotcha #2: getUserFromRequest no longer returns a
hardcoded { userId: 1, email: 'admin@tcgvault.com', role: 'admin' }
when the Authorization header is missing or malformed.
lib/permission-middleware.js
- getUserFromRequest now returns null for missing/malformed Bearer
headers. No console.warn, no NODE_ENV gate — the fallback is gone,
period.
- Token-verify path and DB lookup unchanged.
pages/api/auth/verify.js
- No-token branch now returns 401 instead of fetching the seed admin
via `WHERE email = 'admin@tcgvault.com'`. Closes the admin-record-
leak side of the same bypass.
- JWT-verify branch unchanged.
Known follow-up (flagged but NOT addressed in this PR):
pages/api/collections/[identifier]/cards.js POST/PUT/DELETE handlers
dereference user.userId without a null guard. Previously masked by
the synthetic admin (anonymous-write-as-admin on collections owned
by user 1 was the security hole). Now degrades to NPE → 500 instead
of a clean 401. Security is improved either way; cosmetic 500-vs-401
fix lives in a separate one-line follow-up PR.
Convoy: fix-auth-bypass / Brief 2
Co-authored-by: Cursor <cursoragent@cursor.com>
- New `lib/auth-secret.js` is the single source of truth for `JWT_SECRET`
and the canonical `JWT_TOKEN_TTL = '24h'`. Module throws at import time
if `process.env.JWT_SECRET` is unset — no silent fallback to the literal
`'your-secret-key-change-in-production'`.
- 7 callers refactored to import from the helper:
lib/permission-middleware.js
pages/api/auth-utils.js (also drops unused `'7d'` → JWT_TOKEN_TTL)
pages/api/auth/login.js (also routes via auth-utils.generateToken)
pages/api/auth/register.js (same)
pages/api/auth/verify.js (Brief 2 still owns the no-token admin branch)
pages/api/favorites.js
pages/api/users/search.js
- `process.env.JWT_SECRET` now appears exactly once in the JS source
(lib/auth-secret.js). `your-secret-key-change-in-production` is gone.
- TTL drift reconciled: auth-utils used `'7d'`, login/register used
inline `'24h'`. Both now route through imported `JWT_TOKEN_TTL` (24h).
Pre-deploy reminder: Vercel must have `JWT_SECRET` set before merge or
serverless functions refuse to boot. Existing tokens (signed against the
fallback literal) will be invalidated — users will need to log in again.
Resolves AGENTS.md gotcha #3. Brief 2/3/4/5 still pending in convoy.
Convoy: fix-auth-bypass / Brief 1
Co-authored-by: Cursor <cursoragent@cursor.com>
Removes four unauthenticated dev endpoints that were shipped to production:
- pages/api/simple.js (info leak)
- pages/api/test-auth.js (auth diagnostic / token-mint side door)
- pages/api/test-db.js (DB connection diagnostic)
- pages/api/setup-database.js (public POST that ran DDL + seeded admin)
setup-database is the highest-impact removal: it was a public endpoint
that triggered schema bootstrap and seeded the default admin credentials
(admin@tcgvault.com / admin123). AGENTS.md gotcha #5.
Also adds a new `forbidden-endpoints` job to .github/workflows/ci.yml
that fails the build if any of the four deleted paths re-appear OR if
any new pages/api/test-*.js file is added. Cheap insurance against a
future agent re-introducing a dev endpoint from an outdated tutorial.
README: drops the single `GET /api/test-db` line under "Health Check".
Rest of the API list is intentionally left for the doc-writer pass.
Verified locally:
- npm run build exits 0 (no source callers — confirmed via grep across
pages/, components/, lib/)
- CI guard local simulation: clean → OK; with test-fake.js → FAIL; OK
after cleanup
Resolves AGENTS.md gotcha #5. Brief 1/2/4/5 still pending in convoy.
Convoy: fix-auth-bypass / Brief 3
Co-authored-by: Cursor <cursoragent@cursor.com>
Doc-writer pass for convoy bump-next-js (PR #4 / commit e57ea17).
Single file touched: AGENTS.md (+5 / -1).
- § 1 Project overview: Framework line bumped Next.js 15 -> 16, with
a cross-reference to new Gotcha #9 for the typescript-is-just-for-lint
context.
- § 4 Common gotchas: three new entries that future agents need to
know about but wouldn't infer from the code:
- #9: typescript@^5.9.3 is installed purely so eslint-config-next@16's
bundled typescript-eslint chain can satisfy its hard require('typescript')
at module load. No tsconfig.json, no .ts files, no @ts-check. Decision C.
- #10: ESLint pinned to ^9.39.4 (maintenance), not v10 (latest). v10
surfaced Risk R15 empirically (TypeError: scopeManager.addGlobals)
via @typescript-eslint/scope-manager@8.59.4 predating v10 GA.
Do not bump independently — wait for queued bump-eslint-10
follow-up convoy. Decision D.
- #11: Turbopack is now the default bundler in next dev/build.
Fallback per-command is --webpack. Do not pre-emptively switch.
- § 7 Deployment: reference VERCEL_AUTOMATION_BYPASS_SECRET (env var
name only, no value) for the queued adopt-playwright-smoke convoy
to use against protected preview deploys.
CHANGELOG.md / DEVELOPER_CHANGELOG.md not created — those are deferred
to launch-polish per the convoy's roles section.
README.md staleness (line 16 still says "Next.js 15, React 18,
TypeScript") flagged in the PR description but NOT fixed here per the
docs-pass scope. Pickup: launch-polish.
Co-authored-by: Cursor <cursoragent@cursor.com>
Closes P0 ship-blocker #8 from .convoys/ship-readiness.md. Vercel has
been refusing every deployment since 2025-08-01 with "Vulnerable version
of Next.js detected, please update immediately" — this bump clears
that platform gate and unblocks every downstream preview-smoke and
visual-diff gate that depends on a live preview URL.
Changes (per Brief 1 acceptance criteria, all four gate-1 decisions
applied — see .convoys/bump-next-js.md § Decisions for the audit trail):
- next: ^15.4.2 -> ^16.2.6 (resolves next@16.2.6)
- eslint: ^8 -> ^9.39.4 (Decision D fallback; v10 surfaced Risk R15
empirically — @typescript-eslint/scope-manager@8.59.4 bundled by
eslint-config-next@16 doesn't implement v10's new addGlobals API)
- eslint-config-next: 15.4.2 -> ^16.2.6
- typescript: newly added at ^5.9.3 as a devDep (Decision C; required
by typescript-eslint chain regardless of ESLint major)
- scripts.lint: "next lint" -> "eslint ." (next lint removed in 16)
- next.config.js: images.domains -> images.remotePatterns (deprecated
and removed in Next 16; preserves the three CDN hosts Scryfall,
Pokemon TCG, Lorcana API for eventual next/image adoption)
- .eslintrc.json deleted (eslint-config-next@16 is flat-config-only)
- eslint.config.mjs added (verbatim shape from Next docs; verified
forward-compatible with v10 so bump-eslint-10 will not need to
touch this file)
Out of scope (deferred to dedicated convoys):
- React 18 -> 19 (bump-react)
- App Router migration (multi-month effort)
- Test runner adoption (adopt-vitest, adopt-playwright-smoke)
- Lint baseline cleanup (fix-lint-baseline) — new v9 baseline is
128 problems (81 errors, 47 warnings), up from prior ~100 due to
eslint-plugin-react-hooks@7.1.1 + @next/eslint-plugin-next@16.2.6
rule additions
- ESLint v10 adoption (bump-eslint-10) — upstream-blocked on
typescript-eslint shipping a v10-tested release that
eslint-config-next then bundles
- TypeScript 6 adoption (bump-typescript-6) — same upstream block
Local verification:
- npm install: clean, no ERESOLVE warnings
- npm run build: exit 0, Next 16.2.6 (Turbopack), ~1.4s compile,
23 static pages + 47 API routes, no images.domains deprecation
- npm run lint: exit 1, 128 problems, runs to completion (tolerated
by CI's `|| true` wrapper; new baseline for fix-lint-baseline)
Co-authored-by: Cursor <cursoragent@cursor.com>
Randall Stillwell