Skip to content

start-here.md

Latest commit

 

History

History
62 lines (49 loc) · 5.98 KB

start-here.md

File metadata and controls

62 lines (49 loc) · 5.98 KB
Raw

Start here

Use this page to reach the right depth without reading the documentation front to back.

Pick your path

You are… Start with Then use
Adopting or evaluating Adopter path Get started, profiles, and upgrade/rollback
Maintaining AgentFlow Maintainer path Modular architecture, workflow, and ADRs
Authoring a provider Provider author path Provider matrix and authoring contract
Operating an installation Operator path Troubleshooting, Cockpit, and evidence contracts

The model at a glance

flowchart LR
  Need["Need"] --> Issue["Issue / acceptance criteria"]
  Issue --> Roles["Role-based delivery"]
  Roles --> Checks["Validation & review"]
  Checks --> PR["Evidence-backed PR"]
  PR --> FollowUp["Release or follow-up"]
Loading

New-user essentials

You only need four ideas to begin:

  1. AgentFlow wraps your existing project; it does not generate or replace it.
  2. One agent normally carries the work through explicit roles, scaling from a single CLI (S0) to multi-CLI orthogonal routing (S1) or an OpenRig squad (S2) only when your setup benefits from it.
  3. Issues and PRs hold durable evidence; .agent-runs/ remains local scratch.
  4. High-assurance work keeps human review before merge.

Onboarding & configuration funnel

Step in your journey Canonical guide
1. Understand in 5 minutes AgentFlow in 5 minutes — problem, role model, and evidence flow
2. First governed change (6 commands) Get started — inspect, preview, confirm, apply, and verify
3. Agent-guided initial setup Assisted onboarding — copy-paste prompt for any coding agent to run the read-only-first preview
4. Choose substrate (S0/S1/S2) & routing Project setup — single-CLI subagents (S0), multi-CLI Four-Eyes (S1), or OpenRig Lean 4-Seat Squad (S2)
5. Day-2 refinement & MCP maintenance Refine your setup — preview-confirmed updates to posture, branches, checks, and routing

Advanced-user map

Concern Canonical document
Continuous configuration and maintenance Assisted configuration and project setup
Phase transitions and composite Phase 3–5 pass Agent workflow
Platform identity, execution target, and transport Runtime platforms and execution targets
Role routing, effort calibration, and Four-Eyes Agent routing
Portable execution intents and provider inspection Capabilities
Collaboration modes, L2 subagents, and Effort Law Intelligent collaboration
Handover acceptance, rework, and councils Role collaboration
OpenRig Lean 4-Seat Squad (S2 substrate) OpenRig integration
Artifact and lifecycle schemas Evidence contracts and lifecycle boundaries
Extension and adapter distribution Extension packs, default skills, and packaging
Lifecycle roles and configurable methods Lifecycle roles and role methods
Release governance Release versioning and publishing
Evals and metrics Agent evals and outcome metrics

For the complete categorized inventory, use the documentation index.