sax3l/accounted
3
0
Fork 0
Code Issues 1 Pull Requests Actions Packages Projects Releases Wiki Activity
1,508 Commits 6 Branches 0 Tags
c0ecf2fa3bebd46bdfd0169efd73b89653d1dfed
Commit Graph

7 Commits

Author SHA1 Message Date
Jakob Wennberg 9fe37b85b5 feat(agents): per-key approval authority, the amount an agent may post unattended (#2079) 
* feat(agents): per-key approval authority, the amount an agent may post unattended

An API key gets an optional ceiling in SEK. Above it the agent may still stage
the work, it just may not finish it alone: a human approves the same verifikat
in the app. Default is NULL, so every existing key keeps its behaviour and
turning this on is entirely opt-in.

Enforced at the two places an API key reaches the ledger, and at both the
refusal happens BEFORE the point of no return:

- MCP: in commitPendingOperation, before the atomic claim, so the operation
  stays 'pending'. Behind the claim it would be caught by the generic handler,
  marked terminal 'rejected', and the staged verifikat would be gone.
- REST: in journal-entries.commit, before commitEntry, so the draft stays a
  draft and the voucher sequence never advances (BFL 5 kap. 7 §). The dry run
  refuses too, rather than promising a voucher number the key cannot deliver.

Not enforced inside commit_journal_entry: a RAISE there is swallowed by
engine.ts into a retryable 500, and it would cost a DROP+CREATE on the function
that issues every voucher number.

Operations whose amount is only known during dispatch (batch allocation, bulk
booking, the settlement link paths) fail OPEN behind an explicit allowlist.
Pricing them ahead of dispatch would be a guess, and a wrong guess silently
breaks batch allocation the day someone sets a limit. The allowlist is derived
from what production actually stores: create_voucher carries total_debit on
1389 of 1389 rows, categorize_transaction carries amount on 2002 of 2003,
create_supplier_invoice_from_inbox carries total on 208 of 228.

This is a blast-radius cap, not a security boundary. A per-entry ceiling is
defeated by splitting one entry into several, and an LLM will find that, so
UNATTENDED_COMMIT_LIMIT_EXCEEDED forbids splitting first: one affärshändelse is
one verifikat (BFL 5 kap. 6 §). A cumulative rolling-window limit is the
primitive that actually bounds exposure and is left to a separate change.

The guard is written NULL-first everywhere. An absent, unparseable or
non-positive ceiling always means unlimited, never "block everything".

Agents read their own ceiling from gnubok_get_agent_briefing instead of
discovering it by burning a staged verifikat on a 403.

Changing a ceiling is auditable: it now renders in behandlingshistorik
(BFL 5 kap. 11 §). The audit trigger already fired on the column, but the
report dropped the event because the field was not in its diff map.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

* chore(skill): regenerate accounted-api skill for the new commit pitfall

apiskill:check is a ratchet: the generated reference must match the endpoint
registry. Never hand-edited.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

* test(agents): pin the DB default itself, and declare the briefing field required

Two review findings, both real:

- the default test stored an explicit NULL, so it stayed green even if the
  column default changed to a positive ceiling: the one change that would
  silently start blocking every existing key. It now omits the column.
- gnubok_get_agent_briefing documents unattended_commit_limit as always
  present and emits it unconditionally, so it belongs in the output schema's
  required list.

Declined the NOT VALID constraint suggestion, with the reason recorded in the
migration: api_keys is 388 rows / 768 kB in production.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

* docs(agents): name the TOCTOU window in the REST ceiling check

A security scan flagged that the line sum is read before commitEntry, so a
concurrent write to the draft's lines can post over the ceiling. Real, and
accepted: closing it means enforcing inside commit_journal_entry, where a RAISE
becomes a retryable 500 and destroys the staged operation on the MCP path.

Recorded in the code rather than left implicit, so nobody later mistakes this
for a hard control. A per-entry ceiling is already defeated by splitting, which
needs no race; the cumulative rolling-window limit is the primitive that bounds
exposure.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

* fix(agents): price the settlement and batch paths that were bypassing the ceiling

A security scan flagged that known money-posting operations fail open, and it
was right. The first cut priced only create_voucher, categorize_transaction and
create_supplier_invoice_from_inbox, on the belief that the batch and settlement
paths computed their totals only inside SQL at dispatch. Production says
otherwise: the staged preview already carries the amount, because it is the
number a human is shown when approving the operation.

Over the last 120 days each of these is present and numeric on 100% of that
type's staged rows:

  link_transaction_journal_entry  transaction_amount  1369 rows
  bulk_book_transactions          tx_sum               273 rows
  link_supplier_invoice_voucher   payment_amount        55 rows
  match_batch_allocate            total_allocated       24 rows
  mark_invoice_paid               total                  3 rows

So a key with a ceiling could post any amount through the four largest
settlement paths. Now priced, and the ceiling applies.

Only reconciliation_match stays unpriced: it carries pair_count, which is a
COUNT. Pricing off that would compare pairs against kronor, which is worse than
not enforcing. link_document_to_voucher and attach_document_to_transaction move
no money at all; the transaction_amount they carry is context, not a posting.

Genuinely unpriceable types still fail OPEN. This control can only ever narrow
what a key does, and a wrong guess at an amount blocks a legitimate commit, so
guessing high would leave an agent unable to work.

Adds a test that walks the whole allowlist, so a typo'd field name cannot
silently make a type unpriceable again: that is exactly the hole this closes.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

* fix(mcp): drop the ceiling from the agent briefing, the payload budget has no room

The tools/list context-budget bench sits at 65 000 tokens and main now leaves
roughly 20 tokens of headroom. An always-present field on the briefing's output
schema costs about 85, so this addition alone pushed the bench red.

The bench's own note is explicit that the answer is to demote a tool rather than
raise the ceiling, so raising it here would be the wrong trade for a
nice-to-have.

Nothing is lost that matters: the operation is never destroyed when it is
refused, so discovering the ceiling from UNATTENDED_COMMIT_LIMIT_EXCEEDED costs
one round trip and no work. That error already carries both attempted and limit,
and GET /api/settings/api-keys returns the value. Re-exposing it on the briefing
is worth doing once there is budget to spend.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

* docs(api): spell affärshändelse correctly in the commit pitfall

Fixed in the route's registerEndpoint pitfalls, which is the source; the skill
reference is regenerated from it and never hand-edited.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Jakob Wennberg <311770904+jakobwennberg-oss@users.noreply.github.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
 2026-08-31 15:53:53 +01:00
Jakob Wennberg a717f03898 feat(mcp-oauth): let an account with no company connect and sign up from the OAuth popup (#1814 PR 1) (#1855) 
* feat(mcp-oauth): let an account with no company connect and sign up from the OAuth popup

Identity unlock for agent-first onboarding (#1814, shape B+). A person
with no Accounted account can now connect from an MCP client, create the
account inside the Connect popup and finish the OAuth dance.

- authorize/token no longer require a company: consent renders a
  companyless variant and the key is minted with company_id NULL.
- validateApiKey returns companyId string|null and binds an unbound key
  to the user's first company on the first validation after it exists.
- MCP server: company-dependent tools and data resources answer with a
  structured NO_COMPANY_YET error; the company-independent tools still
  run; telemetry skips when there is no company scope.
- /api/events fails closed instead of throwing for an unbound key.
- authorize forces TOTP enrollment (not just verification) for password
  accounts with no factor, since the middleware skips enrollment for
  zero-company users; BankID-linked accounts stay exempt.
- /login forwards next to /register; register, GoogleAuthButton and
  /auth/callback carry it back to the consent page (callback honours
  only /api/mcp-oauth/authorize, via safeReturnTo); /mfa/enroll
  hard-navigates to /api/* destinations like /mfa/verify.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018wCdzRTatKiDByKB8hCNT6

* refactor(company): move getActiveCompanyId out of the next/headers module

lib/auth/api-keys.ts needs the resolver for unbound-key binding, but
lib/company/context.ts imports next/headers for the legacy company cookie
and Turbopack refuses that import on some of api-keys' import paths (the
preview build failed). The resolver and CompanyContextError now live in
lib/company/active-company.ts; context.ts re-exports them so every caller
and test mock is unchanged.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018wCdzRTatKiDByKB8hCNT6

* fix(mcp-oauth): fail closed on a failed assurance lookup; enroll Back aborts instead of looping

Review findings on #1855: requireAal2 let consent through at AAL1 when
getAuthenticatorAssuranceLevel() returned nothing and a verified factor
existed. Only a positive AAL2 answer passes now; a failed lookup and the
inconsistent verified-factor-at-AAL1 case both step up to /mfa/verify.
Back on /mfa/enroll with the consent page as returnTo went straight back
into the redirect loop; it now aborts to the app.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018wCdzRTatKiDByKB8hCNT6

---------

Co-authored-by: Jakob Wennberg <311770904+jakobwennberg-oss@users.noreply.github.com>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
 2026-08-25 11:25:42 +02:00
Jakob Wennberg ec27228a8e style: remove em/en dashes repo-wide, add CLAUDE.md rule against them (#890) 
Em dashes (—) and en dashes (–) had spread across comments, docs, tests,
and a few UI strings, reading as AI-generated boilerplate rather than
house style. Replaced each with punctuation matching its context: colon
for explanatory clauses, comma for asides, plain hyphen for numeric/legal
ranges (e.g. "21-23§"), "to"/"till" for date ranges, parentheses for
paired-dash asides. messages/en.json and messages/sv.json were fixed by
hand together to keep sv/en in sync.

Left untouched where the dash is the functional subject rather than
decorative punctuation: date-range-parser.ts's separator regex,
charset-repair.ts's CP1252 byte-mapping table (and its test), the SIE
encoding mojibake docs, generic-csv.ts's minus-sign normalizer, the
agent system-prompt files that already instruct against em dashes, and
a golden iXBRL test fixture compared byte-for-byte.

Also fixes two bugs surfaced along the way: an off-by-one in
ApiKeysPanel's scope-label split (a leftover from an earlier partial
pass), and a charset-repair test that had lost the literal en-dash it
exists to verify.

Regenerated the agent atom seed migration (skills:generate) since 27
SKILL.md files changed. Added a CLAUDE.md rule against em/en dashes,
with an explicit carve-out for the functional-dash cases above.

Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
 2026-07-04 15:58:06 +02:00
Jakob Wennberg 0bc81d4c88 feat(auth): SoD acknowledge on stage+approve keys + agent:write scope for memory tools (P0-3) (#681) 
* feat(auth): SoD acknowledge on stage+approve keys + agent:write scope for memory tools

Segregation of duties on API keys is now warn + explicit acknowledgement
(not block): minting a key with any staging write scope AND
pending_operations:approve returns 409 API_KEY_SOD_CONFLICT unless the
caller re-POSTs with acknowledge_sod: true. The acknowledgement is recorded
(sod_acknowledged_at / sod_acknowledged_by) for an auditable risk acceptance
(ISO 27001:2022 A.5.3 / BFNAR 2013:2). The create UI surfaces an inline
warning and an explicit confirm dialog before submitting the ack — the
default "all scopes ticked" create routes through that path.

Also introduces the agent:write scope and maps the previously-UNMAPPED memory
tools gnubok_remember_fact / gnubok_forget_fact to it. Because unmapped tools
were callable by any key, the migration grandfathers agent:write onto every
existing non-revoked key with an explicit scope list so nothing regresses;
new keys must opt in. agent:write is deliberately excluded from the default
grants and is NOT a staging scope (no SoD conflict with approve).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(db): enforce both-or-neither on the SoD acknowledgement pair

Review finding (Greptile P2): sod_acknowledged_at/sod_acknowledged_by were
independently nullable, so a partial write could silently pass and undermine
the auditable risk acceptance (ISO 27001 A.5.3 / SOC 2 CC6.1). Adds a
paired-NULL CHECK constraint + pg-real coverage for both partial-write
directions.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* docs(auth)+feat(auth): compliance-review round — self-attestation documented, ack logged, SoD boundary assumption captured

- Migration header now states explicitly that the SoD acknowledgement is a
  SELF-attestation by deliberate design (enskild firma has no second person;
  the claude.ai approval flow needs stage+approve on one credential) — the
  control objective is informed consent + audit record, not dual control.
- The acknowledge_sod=true path now emits a structured log.warn
  (api_key.sod_acknowledged with key id/prefix, conflicting scope, scopes,
  acknowledger, company) so the acceptance lands in the logging pipeline in
  addition to the sod_acknowledged_* columns (ASVS V16.1.1).
- STAGING_SCOPES carries the documented system control (BFNAR 2013:2
  systemdokumentation) for why agent:write is not a staging scope: memory
  tools write advisory agent context and cannot stage räkenskapsinformation.

Dismissed as by-design/verified: hard-block and second-approver remediations
(user decision: warn + acknowledge); scope-update gap (the [id] route only
supports DELETE — scopes are immutable post-creation); session-auth concern
(withRouteContext is cookie+MFA only; API-key auth exists only on /api/v1
and MCP).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* chore: re-trigger CI (Supabase Preview 502 infra hiccup)

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
 2026-06-06 11:13:35 +02:00
Jakob Wennberg db592d922d feat(api): v1 REST API foundation — auth wrapper, scopes, registry, smoke endpoints (#450) 
* feat(api): v1 REST API foundation — auth wrapper, scopes, registry, smoke endpoints

Lay the substrate for the public REST API at /api/v1/*: Bearer-auth wrapper
that reuses the existing api_keys + idempotency machinery, an extended scope
catalogue (companies, events, webhooks, operations, documents, compliance),
v1 response envelopes (data + meta with request_id, api_version, audit block,
cursor pagination), an error envelope with recovery_hint / docs_url /
valid_alternatives derived from the existing structured-error registry, and
a Zod schema registry that generates the OpenAPI 3.1 spec with x-action-risk
/ x-idempotent / x-reversible / x-dry-run-supported extensions.

Ships discovery routes (/llms.txt, /.well-known/skills/index.json) and three
smoke endpoints (GET /api/v1/health, /api/v1/companies, /api/v1/openapi.json)
so the wrapper is exercised end-to-end. Includes the api_keys.mode (test|live)
migration and 41 unit tests covering auth, scope, company-membership,
idempotency replay, dry-run, pagination, response shape, and scope resolution.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(api): harden v1 foundation — cursor validation, security headers, forensic logs

Address compliance-swarm findings on PR #450:

- OWASP V2.3: decodeDefaultCursor now validates the cursor's ts as ISO 8601
  and id as UUID. A crafted cursor previously could inject untyped strings
  into a query's .gt(field, value); PostgREST would have rejected them, but
  validating here keeps the failure mode predictable (stale cursor → reset)
  rather than 400-ing.
- OWASP V3.4: public discovery routes (llms.txt, .well-known/skills,
  openapi.json) now stamp X-Content-Type-Options: nosniff, Referrer-Policy,
  X-Frame-Options: DENY. New lib/api/v1/security-headers.ts helper.
- OWASP V16: security event logs (missing token, validation failure,
  insufficient scope, company-membership deny) now include source IP
  (x-forwarded-for / x-real-ip) and User-Agent for forensic correlation.
- OWASP V8.2.1 / ISO A.8.3: GET /api/v1/companies emits a warn log when the
  PostgREST archived_at filter unexpectedly returns a row with a null
  company join, surfacing silent data-integrity regressions instead of
  hiding them behind the existing pickCompany() === null filter.

Pushing back on (not changed):
- GDPR Art.32 cursor HMAC signing — cursors only paginate within a user's
  own user_id scope; cross-tenant probe surface doesn't exist yet.
- GDPR Art.25 org_number in list — Bolagsverket public-record data, removing
  forces N+1 fetches to make the response useful.
- SOC 2 CC6.3 service-role bypasses RLS — defense-in-depth IS the design;
  the wrapper's company_members membership check is the technical control.
- ISO A.8.12 public OpenAPI spec — intentional, mirrors Stripe/Twilio.

5 new pagination tests cover the cursor validators. 46/46 v1 tests pass.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(auth): detect Supabase duplicate-signup obfuscation on register

Supabase obfuscates duplicate signups to prevent user enumeration: when an
email already belongs to a confirmed account, signUp returns data.user with
identities: [] and no error, and sends no email. Without detecting this case
we showed the "check your email" screen to the user, who then waited for a
mail that never arrived.

Detect the empty-identities response and surface it via duplicateEmail state
so the UI can branch on it.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(api): second-pass hardening — CSP, IP truncation, cursor scope comment

Address the second compliance-swarm sweep on PR #450:

- OWASP V3.2: PUBLIC_SECURITY_HEADERS now includes Content-Security-Policy
  default-src 'none'; frame-ancestors 'none'. Free win for JSON/text-only
  public routes (no script, style, image, or form contexts).
- GDPR Art.5(1)(f): truncate IPs before logging — IPv4 to /24, IPv6 to /48.
  Preserves diagnostic value (ASN, abuse-pattern correlation, city-level
  geolocation) while eliminating point-of-presence identification. Standard
  pattern used by Google Analytics anonymize_ip. Exported truncateIp() so
  other surfaces can adopt it.
- OWASP V8.2.1: explicit comment in GET /api/v1/companies documenting that
  the cursor's joined_at is applied AFTER user_id filter, so a tampered
  cursor can only reorder rows the caller already owns. Cursors deliberately
  unsigned; trade-off documented.

Pushing back on second-pass findings (not changed):
- ISO A.8.12 / SOC 2 CC6.3 health/llms.txt/skills exposing service name +
  API version + MCP URL — these are intentional disclosures for a public
  3rd-party developer API; hiding them is theatre.
- GDPR Art.32 logging granted scopes on INSUFFICIENT_SCOPE — diagnostic
  value during incident response outweighs the theoretical privilege-profile
  leak; an attacker who already breached the log store has bigger problems.
- OWASP V2.2 route-level Zod for cursor — decodeDefaultCursor already
  validates strictly; route-level Zod is stylistic.
- GDPR Art.25(2) org_number/entity_type in list — Bolagsverket-public data;
  entity_type materially affects which API calls make sense.
- ISO A.8.15 x-forwarded-for trusted-proxy CIDR — overkill behind Vercel's
  edge which rewrites the leftmost value.

50/50 v1 tests pass (4 new for truncateIp). Build green.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(api): third-pass hardening — Host header injection, anon client, HSTS

Address the third compliance-swarm sweep on PR #450:

- SOC 2 CC6.1 (3× high): llms.txt, openapi.json, and .well-known/skills
  built URLs from the inbound Host header. A spoofed Host could poison
  agent discovery with attacker-controlled endpoints. New
  lib/api/v1/base-url.ts centralises canonical base-URL derivation via
  NEXT_PUBLIC_APP_URL (already a required env var per CLAUDE.md).
- ISO A.8.2 / A.8.5 (2× high): the wrapper's public-scope code path now
  uses an anon-key Supabase client (RLS-respecting) instead of the
  service-role client. A future accidental DB call from a public handler
  is constrained to anon-accessible rows. Least-privilege at the
  infrastructure layer.
- OWASP V3.2 (medium): PUBLIC_SECURITY_HEADERS now includes
  Strict-Transport-Security: max-age=31536000; includeSubDomains.
- GDPR Art.5(1)(f) (medium): truncateIp now logs a warn when a non-empty
  x-forwarded-for / x-real-ip payload fails to parse, surfacing spoofed
  or unexpected proxy values to security monitoring instead of silently
  dropping them. The raw value is never logged.
- CC2.3 (low): llms.txt now links the SECURITY.md disclosure policy with
  the security@arcim.io reporting address so agents have a clear
  responsible-disclosure path.

Pushing back on third-pass findings (not changed):
- Cursor HMAC signing — user_id filter is the authorisation boundary;
  cursor scope is bounded to within-user rows. Documented in code.
- org_number in companies list — Bolagsverket public data; the swarm's
  "could be enskild firma personnummer" framing isn't accurate (enskild
  firma org_number IS the personnummer, but it's already in the public
  Bolagsverket business register).
- Health endpoint information disclosure — intentional for a public
  developer API; matches Stripe/Twilio convention.
- llms.txt / skills index MCP URL disclosure — that's the file's purpose.
- Cache-Control public on discovery routes — content is by definition
  public; getCanonicalBaseUrl() removes the previous spoof concern.
- Duplicate-email screen — user's own input; out of scope for this PR.

50/50 v1 tests pass; @supabase/supabase-js#createClient mocked so the
public-path tests don't need real env vars.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(test): widen validateApiKey result assertions to include mode field

The core-only CI job failed on two pre-existing api-keys.test.ts assertions
that used strict toEqual matching against the old (userId, companyId, scopes)
shape. The wrapper migration in this PR widened that shape with mode,
apiKeyId, and apiKeyName.

Update both existing assertions to match the current shape and add a third
test that exercises the mode='test' path. 3027/3027 vitest tests now pass
locally.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(api): fourth-pass hardening — env guards, IP range check, headers on wrapped routes

Address the fourth compliance-swarm sweep on PR #450:

- ISO A.5.17 / SOC 2 CC6.1 (high): createAnonClient now fails closed with
  an explicit Error if NEXT_PUBLIC_SUPABASE_URL or _ANON_KEY are missing,
  surfacing misconfiguration on the first request instead of throwing
  deeper in the handler with no context.
- GDPR Art.5(1)(f): truncateIp now rejects IPv4 with out-of-range octets
  (>255). '999.999.999.999' now returns undefined instead of a pseudo-IP
  that would pollute abuse-pattern analysis. Edge octets (0, 255) still
  accepted. 2 new tests.
- OWASP V3.2 / V3.3: the wrapper's stampHeaders step now applies the full
  security header set to every wrapped v1 response (CSP, HSTS, X-Frame,
  Referrer-Policy, X-Content-Type-Options) PLUS X-Robots-Tag: noai,
  noimageai so authenticated payloads are excluded from AI training sets.
  Public discovery routes (llms.txt, skills index, openapi.json)
  deliberately omit X-Robots-Tag — being AI-discoverable is the whole
  point of those surfaces.
- New WRAPPED_RESPONSE_HEADERS export separates the two contexts.

Pushing back on:
- SOC 2 CC6.1 medium "API key prefix in public docs aids brute force" —
  inverted logic. Every public API publishes its key prefix specifically
  so secret scanners (GitHub Advanced Security, GitLeaks) can detect
  leaks. Stripe (sk_live_), GitHub (ghp_), OpenAI (sk-) all do this.
- SOC 2 CC6.3 medium "formal risk register for unsigned cursors" — org
  -level documentation, outside this PR. Code-comment already documents
  the trade-off.
- SOC 2 CC2.3 low "llms.txt hardcodes security@arcim.io" — same address
  as SECURITY.md; no drift risk.

Flagged separately (not changed): the register-page duplicate-email
detection in this branch defeats Supabase's user-enumeration obfuscation
(GDPR Art.5(1)(c) × 2, ISO A.8.11). Substantive product decision: UX (no
infinite-wait for non-existent accounts) vs security (no enumeration).
GitHub and Stripe Atlas pick UX; some pick security. Owner's call.

3029/3029 vitest tests pass.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(api): address Greptile review on PR #450

- P1 (companies/route.ts): keyset pagination was missing its tiebreaker.
  The cursor encoded (joined_at, id) but the filter only applied
  .gt('joined_at', ts) — same-joined_at rows on a page boundary could be
  skipped or duplicated. Also the encoded id was companies.id while the
  sort was on company_members, mismatched. Fixed: select + sort + encode
  on company_members.id, apply compound
  joined_at.gt.{ts} OR (joined_at.eq.{ts} AND id.gt.{cursor_id}) via .or().
  Side benefit — eliminates the broken-cursor-on-null-join case (#2)
  because company_members.id is always present, no null guard needed.
- P2 (registry.ts): ZodUnion branch had a dead ternary
  (['x','y','z','w'].length > 0 ? undefined : 'object') that always
  yielded undefined. Removed; emit { oneOf: [...] } without top-level
  type (correct JSON Schema for a union).
- P2 (with-api-v1.ts): public-endpoint path was short-circuiting before
  Bearer-token validation, contradicting the JSDoc and PR description.
  Now opportunistically validates a supplied token for rate-limit
  attribution + key tracking; missing/invalid token silently falls back
  to anon (the route is public by definition, so we don't 401). Two
  new tests cover both branches.

3031/3031 vitest tests pass; build green.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-12 20:47:13 +02:00
Mattsson c03582b5c7 Fix/percistent mcp connection (#392) 
* feat(oauth): add support for refresh tokens in OAuth flow and update database schema

* feat(prompts): add MCP prompts and corresponding functionality for prompt retrieval

* feat(auth): enhance error handling for refresh token operations and validation
 2026-05-05 13:48:53 +02:00
Jakob Wennberg b387a77bfd chore: remove Sentry, consolidate migrations, add test coverage (#244) 
* chore: remove Sentry, consolidate migrations, add test coverage

Remove @sentry/nextjs and all Sentry integration code — error tracking
now handled by Recapt. Consolidate 22 incremental migrations into a
single schema sync migration. Add 6 new test suites (auth, invoice
matching, VAT rules, opening balances) and extend report tests with
edge cases. Update Docker image name to gnubok, sync crontabs and
extension presets, fix CSP missing space, simplify journal entry
missing-document dialog.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix: remove viewer bank import migration never applied to production

20260413150000_viewer_bank_import_permissions.sql (PR #234) was merged
to main but never applied to the production database. It references
current_active_company_id() which does not exist in production either.
This breaks fresh installs and Supabase preview branches because the
migration runs before the consolidated schema sync.

Remove it so the migration chain matches production. The viewer bank
import RLS policies should be re-added in a future migration alongside
the helper functions they depend on.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix: correct delete policies for tables without company_id column

Seven tables in the generic delete-policy loop don't have a direct
company_id column, causing fresh installs to fail with "column
company_id does not exist". Fix by moving them out of the loop:

- invoice_items, journal_entry_lines, receipt_line_items,
  supplier_invoice_items → join through parent table
- extension_toggles, notification_settings, push_subscriptions →
  user-scoped (auth.uid() = user_id)

All policies match their existing production definitions.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
 2026-04-15 10:52:00 +02:00