Completion plan
The original 65 outstanding topics are divided into four passes. All four passes are complete: 85/85 reviewed, with no remaining writing assignments. Each pass follows a connected set of responsibilities and produces a reviewed documentation increment, an updated continuation assignment and a distributable HTML edition. The original 20 reviewed topics retain their declared source and evidence scopes.
| Pass | Topics | Result | Total reviewed after delivery |
|---|---|---|---|
| 1 | 16 | Sources, integrations, AI and background work | 36 / 85 |
| 2 | 19 | Product, architecture and application paths | 55 / 85 |
| 3 | 14 | Finance, imports and information security | 69 / 85 |
| 4 | 16 | Infrastructure, delivery, operations and evidence | 85 / 85 |
Pass 1 connects preserved sources to interpretation, integrations and asynchronous outcomes. Pass 2 reuses the existing access/database/workflow articles while explaining application responsibilities. Pass 3 has fewer topics because financial and import reconciliation need deeper evidence comparisons. Pass 4 closes the runtime, release, diagnosis and evidence contracts. This balances review effort; topic counts alone do not measure complexity.
Pass 1: Sources, integrations, AI and background work
Section titled “Pass 1: Sources, integrations, AI and background work”purpose/ownership: Ownership across system boundaries — Which records and outcomes belong to external systems?architecture/providers: External dependency ownership — Which external records and services does Toolora depend on?documents/originals: Mail originals and attachment evidence — Which system preserves mail originals and their coverage?documents/intake: Document candidates and interpretation — When does a preserved attachment become business data?integrations/ownership: Integration ownership — Which responsibilities remain in Toolora when a shared service is used?integrations/accounting: Accounting and Fortnox — Which accounting responsibilities belong to the connector and the local adapter?integrations/mail: Estate mail archive — What does Toolora read from the mail archive?integrations/recovery: Retries and external outcomes — How can an integration recover without repeating an uncertain action?ai/mcp: MCP tools and responses — Which contract describes the current public MCP interface?ai/context: Context and source evidence — How is source information bounded before model use?ai/proposals: Proposals and human confirmation — When does an AI proposal become an ordinary business record?ai/governance: Usage, audit and feedback — How are tool use and unsupported requests accounted for?background/schedulers: Schedulers and enablement — Where are scheduled jobs registered and enabled?background/outcomes: Intent, processing and receipts — How do queued work and completed outcomes differ?background/changes: Change feeds and checkpoints — How does a caller resume authorized change tracking?background/stalled: Stalled work and diagnosis — What evidence distinguishes an idle job from blocked business work?
Pass 2: Product, architecture and application paths
Section titled “Pass 2: Product, architecture and application paths”purpose/product: Product purpose — Who uses Toolora, and which work does it connect?purpose/actors: Actors and responsibilities — Which actors participate in Toolora workflows?purpose/environment: Current environment boundary — Which shared checks can have real external effects?architecture/data-paths: Two application data paths — When does an operation use RLS or the privileged API?architecture/authorization: Identity, company context and resource authority — Where is caller authority established?architecture/transactions: Transaction and external-effect boundaries — Which changes can be committed atomically?architecture/runtime: Declared runtime topology — Which public entry point serves each component?experience/office: Office workspaces — Which work is organized in the office interface?experience/portals: Customer, worker and subcontractor surfaces — How do the portals differ in scope?experience/editing: Editing and conflict states — How does the interface explain saves and concurrent changes?experience/presentation: Responsive layout and accessibility — Which presentation rules apply across devices and themes?frontend/routes: Routes and role surfaces — Where do application navigation and role-specific pages begin?frontend/cache: Company-scoped queries and cache — How is query state separated between companies?frontend/mutations: Mutations and concurrent edits — How do browser edits reach the owning transaction?frontend/configuration: Configuration and generated types — Which configuration and type information belongs in the frontend?api/policy: Route policy and resource scope — What must a privileged handler verify before business work?api/contracts: Request and response contracts — Where are inputs, outputs and errors defined?api/services: Domain and provider services — How is reusable server behaviour separated from HTTP handling?api/errors: Conflicts and uncertain outcomes — How should a caller interpret a conflict or an uncertain provider result?
Pass 3: Finance, imports and information security
Section titled “Pass 3: Finance, imports and information security”finance/billing-basis: Billing basis and invoices — How is accepted commercial value made available for invoicing?finance/deductions: ROT and RUT allocations — Which facts determine household-work deduction evidence?finance/reconciliation: Payments and reconciliation — What supports a conclusion about payment and outstanding balance?finance/costs: Purchase costs and allocations — How can one supplier cost be attributed to several projects?finance/time-payroll: Time and payroll handoff — Which time is eligible for a payroll handoff?finance/reports: Reports and amount definitions — How do report datasets determine totals and coverage?imports/sources: Preserved sources and identifiers — Which source identity must survive an import?imports/replay: Historical replay and side effects — Why must replayed history avoid forward-looking notifications?imports/reconciliation: Identity, relation and value checks — What must reconciliation establish beyond matching row counts?imports/cutover: Operational transition — When does Toolora become the system of action for new work?security/trust: Trust boundaries and privileged access — Where does an operation cross into a more privileged context?security/credentials: Credentials and configuration — Where do secret values belong, and how are replacements verified?security/data: Personal data, originals and retention — Which data lifecycle rules apply to records and attachments?security/logs: Audit, logs and incident evidence — What can be recorded without copying sensitive content?
Pass 4: Infrastructure, delivery, operations and evidence
Section titled “Pass 4: Infrastructure, delivery, operations and evidence”infrastructure/services: Runtime components and services — Which runtime component serves each application entry point?infrastructure/configuration: Configuration and service identities — How do runtime identities obtain their component configuration?infrastructure/publication: Release directories and publication — How is a built candidate made current?infrastructure/ownership: Shared infrastructure ownership — Which capabilities should Toolora consume rather than reproduce?delivery/local: Local development and synthetic data — How is an isolated development environment prepared?delivery/tests: Test levels and evidence — Which checks run automatically, and what can they prove?delivery/releases: Pull requests and application release — How is a coherent change reviewed and published?delivery/schema: Schema delivery and compatibility — How do database changes remain compatible with application publication?operations/health: Health, readiness and versions — What does each runtime signal establish?operations/diagnosis: Workflow diagnosis — How is an observed failure traced to its owning component or record?operations/recovery: Backup and recovery evidence — What must a recovery check include before an application outcome is claimed?operations/incidents: Incident response — How is an incident contained and its recovery verified?decisions/authority: Document ownership and decisions — Which source owns a rule when documents disagree?decisions/limits: Known limits and open decisions — How are unresolved definitions and missing evidence represented?decisions/status: Implementation and release status — How do implemented, merged, deployed and verified states differ?decisions/evidence: Scenario evidence and coverage — How should a reader interpret a test or delivery receipt?
Completion criteria for every pass
Section titled “Completion criteria for every pass”A reviewed topic answers its registered question with the current mechanism, responsible records/components, admission and failure boundaries, a synthetic example or diagnostic path, and scoped evidence. Preserve article types, topic IDs and old summary anchors. Link the owning explanation instead of repeating whole access or transaction chapters. Split articles when that improves the reader’s route; use a common reviewed source set when several topics share an implementation.
Source review does not establish deployed behavior. Distinguish inspected tests, new executions, reused identical-input evidence and named release observations. A missing source or unresolved product decision remains explicit; a favorable summary is not proof. Keep provider-write, private-data and environment boundaries from AGENTS.md throughout all four passes.
Before delivery, reconcile the register and links, inspect the full diff, run the handbook’s Node 24 verify/browser/package commands and selected repository checks, resolve fresh required CI, and verify the extracted clean HTML edition. Record actual results in the PR. Continue its branch while open; start from main after merge. During the four passes, NEXT.md assigned each complete following pass. After pass 4, NEXT.md defines maintenance, concrete evidence gaps and implementation-driven review rather than another writing batch.
Maintenance after completion
Section titled “Maintenance after completion”The four-pass writing assignment is complete. NEXT.md defines implementation-driven maintenance, canonical-link review and scoped evidence updates. Independent product acceptance and the recorded gaps remain separate; completion of these 85 answers does not close them.