| name | dsg-multi-governance-orchestrator |
| description | Use this skill for DSG multi-source governance work that combines UI trust upgrades, action-layer permission gates, deterministic execution, marketplace/enterprise cutover, portable SaaS architecture |
DSG Multi Governance Orchestrator
This skill merges the five DSG uploaded sources plus the user-supplied deterministic/marketplace patch into one reusable SaaS operating standard.
Source map
skill.zip -> DSG UI / visual system / trust-first product posture.
skill_archive.zip -> DSG Action Layer GED: plan pane, execution pane, permission gate, audit model, runtime state machine, starter app direction.
dsg-agent-os-portable-deploy.zip -> portable SaaS/control-plane Next.js app surface.
dsg-one-agent-ready-autonomy-v3.zip -> DSG ONE dashboard/backend app surface.
dsg_action_layer_gate_debug_report_2026-04-25.md -> NO-GO launch gate evidence: production build/readiness/lockfile must be green.
ข้อความที่วาง (1).txt -> deterministic 10x execution protocol, marketplace acceptance, M1/M2 cutover, and Termux/Codex/Multica bootstrap scripts.
Core rule
Be direct. Protect production. Do not mix demo-only changes with launch readiness.
Default release posture:
- NO-GO when build, lockfile, health, readiness, manifest, or trust pages are missing/failing.
- GO only when all gates are green with current evidence.
- Never claim production readiness from static UI alone.
Deterministic execution protocol
Use this sequence for multi-agent or parallel work:
- T0 Input Lock: freeze schema, code, config, tests, constraints, and desired output.
- T1 Dependency Graph Build: parse work into a DAG and classify independent/dependent/write-conflict domains.
- T2 Parallel Execute: run independent read-only or partitioned-write tasks in isolated scopes.
- T3 Deterministic Merge: normalize outputs, sort by
(stage, topo_index, task_id), merge with explicit conflict log.
- T4 Verification Gate: replay merge, compare hashes, validate dependency order, run correctness checks.
Fail fast if replay hash differs.
Permission and action-layer model
Always separate major work into:
- user-facing plan,
- explicit approval,
- live execution with evidence.
Only checkpoint the user for external login, new app connections, privileged external settings, destructive actions, publishing, sending, payment, booking, billing, or equivalent public actions.
Use only these permission verdicts:
allow
needs user takeover
deny
Production cutover standard
Use only two production milestones:
M1: Production Cutover
M2: Hardening + Launch
Reject mock routes, server-memory source of truth, localStorage persistence, demo-only workflows/actions/pages, and work unrelated to marketplace readiness.
Production acceptance requires DB-backed submit/approve/reject/escalate, server-side org/RBAC, DB-backed dashboard state, complete audit trails, entitlement gates, trust pages, and green health/readiness/smoke checks.
Launch GO/NO-GO gate
For any launch claim, verify in this order:
npm install --package-lock-only
npm ci
npm run typecheck || true
npm run test || true
npm run build
npm run verify:production-manifest || true
./scripts/go-no-go-gate.sh <base-url> || true
Required runtime checks:
/api/health returns HTTP 200 and ok: true.
/api/readiness returns HTTP 200 and ok: true.
/terms, /privacy, /security, /support are reachable.
- Protected routes return expected
401 or 403 when unauthenticated.
Architecture page generation
When asked to add the architecture page without replacing existing pages:
Portable SaaS Next.js app
Create app/dsg-architecture/page.tsx. Do not replace app/page.tsx.
Route: /dsg-architecture
DSG ONE dashboard app
Create frontend/app/dashboard/architecture/page.tsx.
Patch nav in frontend/app/dashboard/layout.tsx with a new Architecture item using Workflow from lucide-react.
Route: /dashboard/architecture
UI/UX trust standard
Use references in references/multi-source-ui-action-reference.md and launch gates in references/launch-gate-debug-report-2026-04-25.md.
Design direction: premium, credible, enterprise-safe, futuristic control-room/NASA energy, readable on mobile and desktop, trust-first, task-first, low-friction.
Avoid touching auth, billing, DB, runtime gates, production env, or deployment config unless explicitly requested.
Output format
## What I changed / would add
[direct summary]
## Files
[file paths]
## Code
[copy/paste code or commands]
## Validation
[commands]
## GO/NO-GO
[current verdict]
Included references and assets
references/source-inventory.md maps the 5 uploaded files plus the patch file.
references/multi-source-ui-action-reference.md contains the UI, action-layer, deterministic, marketplace, and cutover operating standards.
references/launch-gate-debug-report-2026-04-25.md contains launch readiness evidence and blocker rules.
scripts/apply-architecture-page.sh contains a copy/paste helper for adding architecture pages to supported repo layouts.