Jakob WennbergandClaude Opus 4.8
bc61862e76
* feat(agent): telemetry completeness + durability, CI gates, commit_method provenance Quick wins from the "Building AI systems that ship" audit: - mcp.tool_called gains errorMessage (message_sv, truncated 500 chars) on all failure exits; new mcp.skill_loaded event on every gnubok_load_skill (all tiers) so atom usage is finally measurable - event_log: (event_type, created_at) index; cleanup cron keeps mcp.*/agent.* telemetry 180 days (delivery events stay 30) - CI: lint ratchet (npm run check:lint — 60 legacy errors baselined, fails only on NEW errors) and a pg-real coverage gate (migrations touching trigger/RPC/RLS/DEFERRABLE require a *.pg.test.ts change; escape hatch: -- pg-test: covered-by/skip) - journal_entries.commit_method CHECK widened with 'api_key'/'agent'; the MCP approve path records 'api_key' truthfully instead of 'user_accept' (agent_first_vision §8 P0-1). 'agent' is reserved — ALL MCP traffic (incl. claude.ai OAuth, whose access_token is a minted API key) authenticates as api_key today Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * feat(import): derive opening balances from prior-year #UB when SIE lacks #IB (#675) SIE files exported without #IB 0 rows (only #UB -1) previously imported with zero opening balances. getEffectiveOpeningBalances() now derives IB from prior-year UB for balance-sheet accounts when explicit #IB is absent, surfaces the derivation as an info issue in the import preview, and excludes share-capital vouchers from opening-balance detection. Detection regexes are shared between parser and importer so the two checks cannot drift. 507 lib/import tests pass. (Authored in a parallel session in this checkout; included per request.) Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(review): address PR #677 bot findings — RoPA entry, execFileSync, gate scope note Triage of the compliance-swarm + Greptile findings: Applied: - .compliance/ropa.yaml: new mcp.telemetry processing activity declaring the 180-day mcp.*/agent.* retention, lawful basis, data categories, and the no-args/no-results minimisation (ISO A.8.10, GDPR Art.5(1)(c) — the retention split is now formally documented, referenced from the cron) - check-pg-test-coverage.mjs: execFileSync with argv array — no shell, so a hostile base-ref can't inject (ASVS V13.2.1); verified an injection attempt exits 2 without executing - check-pg-test-coverage.mjs: documented the PR-level (not per-migration) scope of the gate so reviewers know to check coverage per migration when a PR carries several risky migrations (Greptile P2) Acknowledged, no change: - errorMessage PII risk: messages are domain-mapped strings; event_log already persists far richer delivery payloads under the same RLS; now declared in ropa.yaml - cron error envelope: errorResponse maps to the canonical safe envelope and the endpoint is CRON_SECRET-gated - two-pass delete "partial state": TTL deletes are idempotent — the next daily run sweeps whatever a failed pass left behind - skill_loaded actorLabel/sessionId: mirrors the pre-existing mcp.tool_called payload; sessionId is the join key the analytics exist for Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
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: falseon everyinputSchema. Guarded bystrict-schemas.test.ts. Forces clear rejections on hallucinated fields instead of silent ignores.- Descriptions ≤ 280 chars. Guarded by
output-schema.test.ts. NoArgs:/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?. Thestaged: trueboolean is the explicit completion signal; agents must not infer completion from prose. Do NOT introduce a parallel{ success, shouldContinue, output }envelope. period_statusthreading — any tool that ties to a fiscal-period-bound date (categorize, mark paid, create voucher, correct/reverse entry, approve supplier invoice) passesdateForPeriodChecktostagePendingOperation. Response then includesperiod_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.tsTOOL_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 tolib/pending-operations/__tests__/voucher-executors.test.tsif the tool stages a newoperation_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 dispatchertool-result.ts—withNext(),toToolError()response helpersresources/— read-onlyAccounted://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 promptsskills/— domain-knowledge skill bodies served viagnubok_load_skill__tests__/— strictness guards + per-tool coverage