Skip to content

README.md

Latest commit

 

History

History
71 lines (56 loc) · 3.42 KB

README.md

File metadata and controls

71 lines (56 loc) · 3.42 KB

ControlFlow Tutorial (English)

A practical guide to the ControlFlow multi-agent orchestration system. Suitable for both newcomers and developers who want to understand the system deeply.

Table of Contents

# Chapter Topic
00 Introduction What ControlFlow is and what it delivers
01 Quick Start Orientation in 30 minutes
02 Architecture Overview Mental model of the whole system
03 Agent Roster All 13 agents — roles, inputs, outputs
04 P.A.R.T. Specification Mandatory structure of every agent file
05 Orchestration How the Orchestrator governs the process
06 Planning How the Planner turns ideas into plans
07 Review Pipeline Adversarial PLAN_REVIEW before execution
08 Execution Pipeline Phases, waves, quality gates
09 Schemas (Contracts) All 15 JSON schemas — purpose and key fields
10 Governance 5 governance files — permissions and runtime knobs
11 Skills (Patterns) Reusable expert patterns for agents
12 Memory Architecture Three-layer memory model
13 Failure Taxonomy 4 failure classes and routing
14 Eval Harness Offline validation suite
15 Case Studies End-to-end scenario walkthroughs
16 Exercises Practice tasks by level
17 Glossary Key terms with chapter references
18 FAQ Frequently asked questions

Reading Trajectories

🟢 New to the system

00 → 01 → 02 → 03 → 04

🟡 Understanding orchestration and planning

05 → 06 → 07 → 08

🔵 Infrastructure: schemas, governance, skills, memory, evals

09 → 10 → 11 → 12 → 14

🔴 Practice

13 → 15 → 16 → 17 → 18

Chapter Template

Each chapter follows this structure:

  • Why this chapter — what you will understand after reading.
  • Key concepts — definitions of terms introduced in the chapter.
  • Mermaid diagram — visual model of the described process or structure.
  • Detailed text — explanation with examples and code references.
  • Common mistakes — what is misunderstood most often.
  • Exercises — practice tasks (🟢 beginner / 🟡 intermediate / 🔴 advanced).
  • Review questions — self-check.
  • See also — links to related chapters and files.

Canonical Sources

All chapter content is derived from:

  • Agent files (*.agent.md) — authoritative for agent behavior.
  • governance/runtime-policy.json — authoritative for thresholds, tiers, retry budgets.
  • schemas/*.json — authoritative for inter-agent contracts.
  • docs/agent-engineering/ — authoritative for engineering policies.
  • plans/project-context.md — authoritative for agent roster and conventions.

When the tutorial conflicts with a canonical source, the canonical source wins.

Text Conventions

  • monospace — file paths, field names, enum values, commands.
  • Italic — emphasis on a key point on first introduction.
  • → — "see also" link within or across chapters.
  • Technical terms (agent names, file paths, field names) are in English throughout.