SSLZ Agent documentation · Launch the agent · Classic direct-operator site
SSLZ Agent
Founder-guided Azure readiness, planning, evidence, and approval boundaries
The SSLZ Agent guides founders from “what are you building?” to a reviewable Azure readiness and landing-zone plan. It reuses local scripts, machine-readable contracts, deterministic planning, and narrowly approval-gated integrations. It does not replace the classic Bicep or Terraform Quick Start.
For an underspecified Azure adoption request, it summarizes knowns and unknowns, makes a conditional architecture recommendation suited to the organization and workload, and identifies the next decision before selecting a migration, coexistence, or existing-Azure improvement path. Unknown service fit, account readiness, availability, quota, capacity, billing, and support are never inferred from the prompt.
The launch page uses the profile and implementation from this repository’s main branch. Initial planning requires Git and
GitHub Copilot CLI authenticated with a GitHub account that has Copilot access. It does not require Azure CLI, an Azure
subscription, or Azure login. Node.js is required when you run local planners or tests; Azure inspection and preview
have additional tooling and separate approval requirements.
Not an autonomous hosted bot: SSLZ does not embed a chatbot, run a hosted agent, capture your credentials, or make unattended cloud decisions. Operators choose inputs, review artifacts, provide Azure authentication only when needed, and retain responsibility for deployment, validation, and rollback.
Choose the right path
| Experience | Best for | Execution model |
|---|---|---|
| Classic SSLZ | Teams that want the original one-hour landing-zone path | Direct operator Bicep or Terraform |
| SSLZ Agent | Founders who want guided discovery, plans, evidence, and explicit approval boundaries | Local Copilot CLI plus repository scripts and contracts; writes are absent, disabled, or narrowly approval-gated as documented |
Flow, contracts, and evidence
- Startup agent flow explains account discovery, workload selection, regional planning, IaC review, and execution boundaries.
- Preflight result contract defines account, subscription, billing-topology, check, action, and deployment-plan results.
- Implementation and evidence status is authoritative for what is delivered, what can write, and whether evidence is synthetic, historical, or live.
- Implementation plan records the phased program and review gates.
- Recommendation grounding explains how Azure guidance is checked against Microsoft Learn and kept distinct from the capabilities implemented in this repository.
Planning and readiness
| Area | Documentation |
|---|---|
| Workload selection | Startup workload profiles |
| Region and recovery posture | Hot/Cool regional topology |
| Bicep and Terraform review artifacts | IaC plan generation |
| Privacy-preserving readiness bindings | Readiness evidence |
| Provider registration | Approved provider remediation |
| Baseline deployment boundary | Approved deployment integration |
Migration and multi-environment programs
These programs remain separate from primary-baseline authority. Their lineage does not silently grant execution rights.
- Program lineage envelope
- PostgreSQL migration planning
- Container image and CI/CD migration planning
- Dual-cloud connectivity, DNS, identity, and egress planning
- Control-plane ownership planning
Safety and execution boundaries
- Prefer Microsoft Learn MCP when available; use the approved official-page fetch fallback when it is not. Cite only retrieved content and distinguish documentation from the SSLZ implementation and tenant-specific evidence.
- Keep normal Copilot permission prompts enabled and approve only the proposed tool, command, path, or official URL.
Do not use blanket approval or
--allow-all. - Read-only planners do not log into Azure unless a documented preview or inspection explicitly requires authentication.
- Planning-only migration, image, connectivity, control-plane, and Hot/Cool artifacts do not authorize execution.
- Provider registration and primary-baseline integration accept only reviewed, digest-bound, single-purpose approvals.
- A passing schema or fixture test is synthetic evidence, not proof of a live deployment, migration, failover, image promotion, private connection, capacity allocation, or application-health outcome.
- Follow the implementation status matrix for the remaining human review and live validation gates.
Local validation includes conversational scenarios and synthetic planner tests. A fixture replay through the
v3 planner uses a fixed test clock and synthetic readiness and Defender decisions; it does not establish current tenant
readiness. Capturing JSON from a CLI response and running the planner externally does not demonstrate autonomous
parameter generation or planner execution by the agent. For repeatable use, record the exact main commit you reviewed.
Repository assets
- Agent check catalog and schemas are published under this section; start with the greenfield journey report schema.
- Sanitized contract examples show expected shapes without tenant secrets.
- Planning and validation scripts are source-controlled and run locally.
- Agent contract source contains every schema, profile, example, and check.