Roadmap: What's Built & What's Coming
SGD is being built in phases. Each phase delivers a usable increment — you do not need to wait for the full vision to get value.
Current focus: Coherence & Drift Intelligence (cross-artefact checks, drift detection, alerts) and Cortex (Layer 8 persistent memory — the @repo-sentry/cortex-mcp server is published). The foundation, governance engine, and AI-context layers are shipped.
Phase 1: Foundation Shipped
The core building blocks that everything else depends on.
| Feature | Status | Description |
|---|---|---|
| Capability Model | ✅ Shipped | Hierarchical L1/L2/L3 model mapping business capabilities to repositories (auto-inferred + editable) |
| Terminology Rebrand | ✅ Shipped | "RepoSentry compliance" → "SGD governance" across the platform |
| Feature Spec Registry | ✅ Shipped | Create, version, and approve feature specifications linked to capabilities |
Phase 2: Governance Engine Shipped
The checks and scoring that make governance automatic.
| Feature | Status | Description |
|---|---|---|
| Design System Enforcement | ✅ Shipped | Canonical components, banned patterns, and design tokens — checked on every PR |
| Impact Analysis DAG | ✅ Shipped | Directed acyclic graph of repository dependencies, API contracts, and spec relationships |
| PR Governance Checks | ✅ Shipped | Spec traceability and design-system compliance as GitHub Check Runs |
| Coherence Scoring | ✅ Shipped | 0–100 score per repository with configurable weights and graceful degradation |
| Test Generation from Specs | 📋 Planned | Auto-derive executable BDD tests from a feature spec's Gherkin acceptance criteria, so no scenario ships without a test — coverage feeds the built_coverage drift metric |
Phase 3: AI Governance Shipped
Context generation and AI tool management.
| Feature | Status | Description |
|---|---|---|
| CLAUDE.md Generator | ✅ Shipped | Auto-generate and maintain CLAUDE.md files from the capability model, specs, and ADRs |
| MCP Server | ✅ Shipped | Model Context Protocol server providing specs, ADRs, and governance context to AI tools — @repo-sentry/cortex-mcp is published to npm |
Phase 6: Vision & Stakeholder Governance — Layer 0 Shipped
Govern the shared intent that every other layer is judged against.
| Feature | Status | Description |
|---|---|---|
| Vision Governance Check | ✅ Shipped | Verify every governed repo has a docs/vision.md with the required sections — vision, non-goals, MVP framing, and success measures |
| Spec / ADR Vision Citation | 🛠️ Building | Check that feature specs cite a success measure they move, and ADRs cite the vision element they protect |
| Stakeholder Questions Registry | 📋 Planned | First-class QD-NN decision artefacts linking each business answer to the spec and commit that actions it — the audit chain regulators ask for |
| Orphan-Question Coherence Check | 📋 Planned | Flag questions raised but unanswered beyond a threshold, and specs whose open questions resolve to no QD-NN |
Phase 7: Coherence & Drift Intelligence — Layer 7 Building Now
Catch the slow "boiling-frog" drift that no single PR is responsible for. Active development.
| Feature | Status | Description |
|---|---|---|
| Journey Derivation Engine | ✅ Shipped | Generate the end-to-end user journey from specs and the capability model, and keep it fresh as specs change |
| Cross-Artefact Coherence Checks | 🛠️ Building | Detect contradictions between specs, ADRs, the capability model, and the code — at PR time (SGD-027 checks C1–C9 landing) |
| SGD Drift Detection | ✅ Shipped | A daily job tracking seven coherence metrics per product: spec coverage, built coverage, journey freshness, orphan rate, supersession depth, ADR cadence, and cross-repo contract drift |
| Composite Coherence Score & Trend Alerts | 🛠️ Building | Roll the metrics into a single 0–100 score with predictive threshold and trajectory alerts — drift spotted before it becomes a crisis; Coherence dashboard tab live |
Phase 8: Cortex — Persistent Memory — Layer 8 Building Now
Make the organisation's learning persist across every session and every repo. Active development.
| Feature | Status | Description |
|---|---|---|
| Cortex Foundation | 🛠️ Building | Per-org evidence-weighted, time-decaying knowledge graph anchored to commits, tests, PRs, and spec files |
| Cortex Auto-Ingestion | 🛠️ Building | GitHub signal pipeline feeding commits, CI results, and PR reviews into the graph automatically |
| Contradiction & Cross-Repo Elevation | 🛠️ Building | Surface conflicting knowledge for human resolution; promote a pattern seen in 3+ repos to an org-wide truth, injected everywhere |
| Cortex MCP Server | ✅ Shipped | Progressive context injection — @repo-sentry/cortex-mcp published to npm; developer enrollment guide available |
| Cortex SGD Integration | 🛠️ Building | Seed the graph from specs at high confidence, score drift, and surface it all on the full dashboard. Enterprise scopes + GDPR purge shipping |
Phase 4: Dashboard & Polish Building Now
The visual layer that makes governance accessible to non-engineers.
| Feature | Status | Description |
|---|---|---|
| Coherence Dashboard | ✅ Shipped | Coherence tab with read API and trend endpoint |
| Capability Tree View | 🛠️ Building | Interactive tree visualisation of the L1/L2/L3 capability model with health indicators |
| Spec Browser | 📋 Planned | Browse, search, and filter feature specs with linked PRs, governance results, and deployment status |
Phase 5: ALM Integration Planned
Connect the first and last links of the traceability chain.
| Feature | Status | Description |
|---|---|---|
| Jira / ADO / Linear Connector | 📋 Planned | Bi-directional sync between ALM tools and the SGD spec registry |
| Full Traceability Chain | 📋 Planned | Complete chain from Jira epic to production deployment with every link stored automatically |
Further out Future
Sequenced but not yet scheduled.
- ROI Metrics Dashboard — Velocity, quality, and coherence metrics with before/after comparison and PDF/CSV export
- Spec Quality Scoring — Automated assessment of spec completeness, clarity, and testability
- AI Provenance Detection — Identify which AI tools authored code using git metadata, branch naming, and MCP session data
- AI Code Review — AI-assisted governance check explanations and fix suggestions
- Figma Plugin — Design system rules synced from Figma component libraries
- Production Feedback Loop — Post-deployment metrics fed back into spec and governance effectiveness scoring
- Compliance Reports — Quarterly governance, monthly coherence, annual AI governance, and ad hoc audit reports
- Multi-Model Governance — Policies and metrics per AI model (Claude Opus vs Sonnet, GPT-4 vs GPT-4o, etc.)
- Developer Effectiveness — Individual and team-level governance adoption metrics (opt-in, privacy-preserving)
How to read this roadmap
Status indicators:
- ✅ Shipped — Built and running in the platform.
- 🛠️ Building — Actively in development now.
- 📋 Planned — Designed and scoped. Scheduled for an upcoming phase.
Phase labels:
- Shipped — Delivered
- Building Now — Current development focus
- Planned — Designed, not yet started
- Future — Vision items, sequence may change
Phases are sequential but overlapping — the layer numbers (0–8) reflect the governance model, not strict build order. Foundation, governance, and AI-context layers shipped first; Coherence and Cortex are the current focus.
Feedback
This roadmap reflects our current priorities. If you have feedback on sequencing, missing features, or use cases we have not considered, open an issue on the GitHub repository or reach out to the team directly.