Engineering Governance Spec (EGS)
Benchmarked against: Anthropic — Agent Intelligence Protocol (governance framework) Version: 1.2 | Authority: Captain-approved Full spec:
docs/EGS-v1.2.md+ Cloud UB ub-5d9f36e55da6
The Engineering Governance Specification is SuperPortia's engineering rulebook. It defines how agents develop, test, deploy, and govern software across the fleet.
ADLC Framework
EGS uses the Agentic Development Lifecycle (ADLC) — replacing traditional SDLC for agent-driven development:
| Phase | What happens | Key gate |
|---|---|---|
| Conceptualize | Vision, requirements, feasibility | Captain approval |
| Architect | Design, tech choices, Pre-Flight Check | 66s Review |
| Develop | Implementation, WO-driven | Code standards |
| Test | Verification, EGS compliance | Acceptance criteria |
| Deploy | Ship to production | Deployment gates |
| Govern | Monitor, maintain, evolve | SRE status |
Core principles
Progressive Disclosure (PD)
Load information in layers, not all at once:
- L1: CLAUDE.md — always loaded, minimal
- L2: Rules (
.claude/rules/) — always loaded, behavioral - L3: Skills (
.claude/skills/) — on demand, task-specific
Roles
| Role | Identity | Responsibilities |
|---|---|---|
| Captain | 夏哥 (Wishia) | Vision, strategy, approvals, reviews |
| Chief Engineer | 小克 (Claude Code) | Architecture, decisions, delegation |
| Strategist | 小西 (Claude Chat) | Strategy analysis, reviews |
| Executor | Various agents | Task execution via WO system |
Chapters — rules and their enforcement
Every EGS chapter defines rules (the "law") and has corresponding enforcement mechanisms (the "execution") documented elsewhere in this site. Together they form the complete developer encyclopedia.
| Chapter | Rules (law) | Enforcement (execution) | Docs link |
|---|---|---|---|
| Ch.2 | Code Standards — naming, safety, testing, 12 sections | Linter config, pre-commit hooks, code review | Content in progress |
| Ch.3 | Development Workflow — ADLC phases, verification matrix | WO system + verification gates | Content in progress |
| Ch.4 | Version Management — semantic versioning, changelog | Git hooks + commit conventions | Content in progress |
| Ch.5 | Pre-Flight Check — 3D risk scoring (0–6) | Pre-Flight Check SOP — agent self-assessment before every task | Pre-Flight Check |
| Ch.6 | Deployment Gates — ship criteria, rollback | Deployment gates + SRE monitoring | Content in progress |
| Ch.7 | Company Constitution — 14 supreme articles | Agent Intelligence Protocol + CLAUDE.md rules + hooks | Constitution |
| Ch.8 | Agent Intelligence Protocol — 8 mandatory behaviors | Boot sequence, session hooks, CLAUDE.md L2 rules | AIP |
| Ch.9 | UB Governance — ingestion language, tags, quality, freshness | MTAAA pipeline + agent checklist + Tech Freshness | UB Governance |
Every row in this table is a capability checkpoint. If the "Enforcement" column has a working link, that capability is built. If it says "Content in progress", we know what to build next. This is the docs site as inspection mirror — its structure drives system development.
Compliance
Every agent must comply with EGS. Non-compliance on critical rules (Constitution, Pre-Flight Check, cost awareness) is classified as a work incident requiring root cause analysis.