methodology

Scaffold a new Transitrix architecture-as-text repository (zoned canon/ + field/ + codex/ layout with assistant-neutral AGENTS.md + transitrix.yaml manifest) and drive a first modelling session — enterprise architecture (BPMN, DGCA / Goals, capability map, process blueprint, action schedule, actions tree, nested blocks, scenarios, products, applications) plus codex artefacts (laws, regulations, policies, internal standards). Use when the user wants to start a new Transitrix repo from scratch, or has just cloned one and wants to author their first notation file with validation.

Help a user formulate an architecture decision and land it as a valid, gated Architecture Decision Record in the repo's per-repo decision log — the input door to the agent-authored decision flow method/03-architecture-decision-log.md specifies but ships no workflow for. Runs a Context → Decision → Consequences interview, allocates the next ADR-NNNN id, validates against scripts/check-adl.mjs, and opens a PR. Never self-accepts a decision and never edits an accepted record's body — course changes are a new record plus a status-pointer flip on the old one (supersession).

Ingest raw source material into the OKF single-repo MVP knowledge store — assess, archive, route to the OKF track (extract drafts → verify → human review → promote to knowledge/) and/or the Canon track (extract Transitrix primitives → PR to canon/), and log every decision.

Read-only health check ("doctor") for a Transitrix adopter repository. Emits a short, DATA-FREE report — methodology version, resolved coverage profile, per-zone and per-TYPE counts, an adoption-level indicator, integrity red flags (invalid IDs, misplaced canon elements, unresolved profile), and a tooling check. Reports aggregates and statuses only — never object ids, names, or contents — so the report is safe to share outside the organisation. Idempotent and non-mutating: it reads the zones and never writes one.

Turn raw organisational material (interviews, policies, org charts, spreadsheets, notes) into Transitrix field artefacts and typed canon candidates at scale, with source-quality scoring and a human review queue. Operates the field→canon derivation pipeline — convert a document, emit a field artefact with provenance and a proposed source_quality, extract typed elements and conservative relations that each cite their field source via derived_from, validate them against the canonical schemas and the adopter's coverage profile, and produce a review queue a human gates before anything is admitted to canon. Never writes canon directly.

Collect regulatory-source changes in practice — the operational counterpart to the methodology's codex / SEGMENT / AMENDMENT notation work. Watches codex sources flagged `monitoring_needed: true`, runs a cheap change-signal gate before any expensive extraction, then on signal-moved runs the SCAN → SEGMENT → CLASSIFY → OUTPUT pass: fetches and snapshots the source, slices it into obligation-bearing SEGMENT field artefacts, classifies each segment as a CONSTRAINT or REQUIREMENT candidate (positive obligation → REQUIREMENT by default), emits an AMENDMENT field artefact when an existing source has drifted, updates the codex `scan` block, and stages everything in a review digest for human admission. Never writes canon. Never auto-admits. Never silently drops.

Produce a compliance report (obligation × subject impact matrix, or per-regime coverage metric) from a Transitrix repository, by conversation. Resolves the report's parameters — from the request, from the view spec's defaults, or from a named saved view-config — materialises a versioned view-config artefact, and renders it through the deterministic `cervin export-compliance` CLI to Markdown or PDF. The skill never computes a matrix itself; rendering stays in the CLI so the same config re-renders identically next quarter. Reports are reproducible and auditable: every report has a committed parameter artefact, and the skill states which spec defaults it assumed.