Files

T

Jakob WennbergandClaude Opus 4.8 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

__tests__

feat(auth): SoD acknowledge on stage+approve keys + agent:write scope for memory tools (P0-3) (#681 )

2026-06-06 11:13:35 +02:00

prompts

Accounted rebrand + swarm-skill cleanup + bank-reconciliation fixes (#643 )

2026-06-03 10:52:01 +02:00

resources

Accounted rebrand + swarm-skill cleanup + bank-reconciliation fixes (#643 )

2026-06-03 10:52:01 +02:00

skills

Accounted rebrand + swarm-skill cleanup + bank-reconciliation fixes (#643 )

2026-06-03 10:52:01 +02:00

widgets

Accounted rebrand + swarm-skill cleanup + bank-reconciliation fixes (#643 )

2026-06-03 10:52:01 +02:00

index.ts

Agent v1 bundle: TIC v2 onboarding, in-app assistant gating, sidebar nav, MCP fixes (#584 )

2026-05-28 11:27:01 +02:00

manifest.json

feat: MCP server, API keys, OAuth, and KPI dashboard (#72 )

2026-03-21 12:57:14 +01:00

README.md

Accounted rebrand + swarm-skill cleanup + bank-reconciliation fixes (#643 )

2026-06-03 10:52:01 +02:00

server.ts

feat(bookkeeping): agent attribution into the immutable ledger layer (P0-1) (#678 )

2026-06-06 10:05:48 +02:00

tool-result.ts

Add/ai native supp (#385 )

2026-05-04 11:12:29 +02:00

README.md

Accounted MCP server

JSON-RPC 2.0 server exposing the Accounted bookkeeping engine to MCP clients (Claude Desktop, Claude Code, etc.). Endpoint: /api/extensions/ext/mcp-server/mcp. OAuth and stdio bridge live alongside the API surface — see app/api/mcp-oauth/ and packages/gnubok-mcp/.

Tool authoring contract

Enforced by tests in __tests__/ — these are not style preferences, they're guard rails.

additionalProperties: false on every inputSchema. Guarded by strict-schemas.test.ts. Forces clear rejections on hallucinated fields instead of silent ignores.
Descriptions ≤ 280 chars. Guarded by output-schema.test.ts. No Args: / Returns: / Examples: prose — those belong in JSON Schema. Use agent-native hints ("Use to…", "Call X first", "HIGH risk").
Staged-operation envelope for write tools — outputSchema: STAGED_OPERATION_SCHEMA (server.ts). Fields: staged, risk_level, actor, message, preview, period_status?, next?. The staged: true boolean is the explicit completion signal; agents must not infer completion from prose. Do NOT introduce a parallel { success, shouldContinue, output } envelope.
period_status threading — any tool that ties to a fiscal-period-bound date (categorize, mark paid, create voucher, correct/reverse entry, approve supplier invoice) passes dateForPeriodCheck to stagePendingOperation. Response then includes period_status: { period_id, status: open|locked|closed, lock_date } so widgets and agents disable writes without round-trips.
Scope mapping — every new tool needs an entry in lib/auth/api-keys.ts TOOL_SCOPE_MAP. Missing entries default to deny.
Tests for new write tools — add staging-gate coverage to __tests__/voucher-tools.test.ts (or a sibling) plus executor coverage to lib/pending-operations/__tests__/voucher-executors.test.ts if the tool stages a new operation_type.

Determinism / cache stability

Tool definitions (name, description, inputSchema, outputSchema, annotations) are declared as static object literals at module load — no timestamps, no UUIDs, no Date/Math.random in the definition layer. This makes the tools/list JSON payload byte-stable across requests, which lets agent-side prompt caches stay warm. Do not introduce per-request non-determinism into the definitions block. Anything time-bound or random belongs inside execute().

For internal Anthropic API usage (today only extensions/general/invoice-inbox/lib/extract-invoice-fields.ts): annotate stable prefixes with cache_control: { type: 'ephemeral' } and log usage.cache_read_input_tokens for hit-ratio observability. The 1h TTL from the agent-native API plan (item 10) requires the direct Anthropic API; Accounted's Bedrock path defaults to a shorter TTL.

Payload-size watchdog

payload-size.bench.test.ts enforces a tools/list JSON payload ceiling (currently 25,000 tokens). If the test fires, the right answer is rarely "raise the ceiling" — instead, trim descriptions or leverage gnubok_search_tools (already deployed; tool definitions can defer to it for discovery rather than enumerating in tools/list).

Where things live

server.ts — the tools array + JSON-RPC dispatcher
tool-result.ts — withNext(), toToolError() response helpers
resources/ — read-only Accounted:// URIs (active company, period, recent activity, capabilities, attention items, voucher gaps, chart of accounts, VAT treatments)
widgets/ — inline HTML widgets (receipt-matcher, vat-review)
prompts/ — slash-command-style prompts
skills/ — domain-knowledge skill bodies served via gnubok_load_skill
__tests__/ — strictness guards + per-tool coverage