Agent Manager Template

A GitHub template repository for teams using AI coding assistants (Claude, Windsurf, Cline, etc.) with structured Kanban workflow governance.

What This Is

This repository provides:

1. Kanban-as-Code — Your project board lives in Features/ folder structure, version-controlled with git
2. Agent Governance — Clear rules for when AI agents write design docs vs code vs tests
3. CSV Sync — Bidirectional sync between folder structure and tasks.csv for Google Sheets/GitHub Projects import
4. Onboarding Scripts — One-command setup to install Kanban rules into agent memory

Who This Is For

If you want to use AI coding assistants (Claude, Windsurf, Cline, etc.) with structured workflow governance, this template is for you.

You will build a team in your IDE of choice with each agent having specific roles and responsibilities and work together to complete tasks in a structured Kanban workflow where you (the human) gatekeep task transitions.

How It Works

Kanban Board as Folders

Features/
├── 1-Backlog/          # Raw ideas
├── 2-HLD/              # High Level Design docs
├── 3-TaskReview/       # Architect sign-off queue
├── 4-Implementation/   # Active coding
├── 5-Testing-Agent/    # Automated test suite
├── 6-Testing-Manual/   # Human QA
├── 7-Verified/         # Ready for PR
├── 8-Review/           # PR open
└── 9-Done/             # Merged & complete

Stories are markdown files that move between folders as they progress.

Full Workflow Diagram

flowchart TD
    subgraph BACKLOG["1 · Backlog  👤 Human"]
        B_in["Feature stub\n─────────────\n· title\n· scope\n· out-of-scope"]
    end

    subgraph HLD["2 · HLD  🤖 Architect Agent"]
        H_in["Feature design\n─────────────\n· C4 L1 Context diagram\n· C4 L2 Container diagram\n· Components & data model\n· Stories decomposed here"]
    end

    subgraph TASKREVIEW["3 · TaskReview  🤖 Architect Agent  →  👤 CTO approves"]
        TR_in["Per-story LLD\n─────────────\n· C4 L3 Component diagram\n· Interface contracts\n· Sequence diagrams\n· Gherkin acceptance criteria"]
    end

    subgraph IMPL["4 · Implementation  🤖 Agent"]
        I_in["Code\n─────────────\n· Tests written first (TDD)\n· Implements LLD contracts\n· Updates Gherkin checklist"]
    end

    subgraph TEST["5/6 · Testing"]
        TA["5 · Testing-Agent  🤖\n─────────────\nUnit · Integration\nAcceptance (Gherkin)\nV-Model: all levels"]
        TM["6 · Testing-Manual  👤\n─────────────\nExploratory · UX\nEdge cases requiring\nhuman judgment"]
    end

    subgraph VERIFY["7 · Verified  👤 Human"]
        V_in["Human approval\n─────────────\nAll tests pass\nAcceptance criteria met"]
    end

    subgraph REVIEW["8 · Review  👤 Human"]
        R_in["PR open\n─────────────\nCode review\nFeedback addressed"]
    end

    subgraph DONE["9 · Done  👤 Human"]
        D_in["PR merged\n─────────────\nFeature complete"]
    end

    BACKLOG -->|"Human moves\nfeature to HLD"| HLD
    HLD -->|"Agent signals ready\nHuman moves"| TASKREVIEW
    TASKREVIEW -->|"CTO commits move"| IMPL
    IMPL -->|"Agent routes"| TA
    IMPL -->|"Agent routes\n(can't automate)"| TM
    TA -->|"All pass"| VERIFY
    TA -->|"Needs human QA"| TM
    TM --> VERIFY
    VERIFY -->|"Human moves"| REVIEW
    REVIEW -->|"PR merged"| DONE

    TA -->|"Tests fail"| IMPL

    style BACKLOG fill:#f5f5f5,stroke:#999
    style HLD fill:#dbeafe,stroke:#3b82f6
    style TASKREVIEW fill:#dbeafe,stroke:#3b82f6
    style IMPL fill:#dcfce7,stroke:#16a34a
    style TEST fill:#fef9c3,stroke:#ca8a04
    style VERIFY fill:#f3e8ff,stroke:#9333ea
    style REVIEW fill:#f3e8ff,stroke:#9333ea
    style DONE fill:#d1fae5,stroke:#059669

Loading

V-Model Alignment

The Kanban left leg (design) maps to the V-Model and its right leg (testing):

DESIGN ──────────────────────────────────────────── TESTING
                                                            
Backlog    Requirements / feature intent    ←→  E2E / Gherkin acceptance tests
HLD        Architecture (C4 L1/L2)         ←→  Container smoke tests
TaskReview Design (C4 L3 / LLD)            ←→  Integration tests
Implement  Code                            ←→  Unit tests

Human vs Agent Responsibilities

Who	Does
Human	Creates feature stubs, moves stories between columns, approves gates, merges PRs
Architect agent	Writes HLD, decomposes stories, writes LLD + Gherkin per story
Implementation agent	Writes tests and code from LLD contracts
Testing agent	Runs all test levels, verifies Gherkin, routes to Verified or Testing-Manual

Agents never move stories. Humans commit all column transitions.

Getting Started

1. Use this template to create your project repo
2. Run onboarding to install Kanban rules:

   ./scripts/windsurf_onboarding.sh

3. Create feature stubs in Features/1-Backlog/{epic}/0001-feature-name.md (stories are written later, as output of HLD)
4. Move stories through columns via git mv and commit

Documentation

• KANBAN.md — Full column definitions and workflow rules
• Features/ — Where you create and move story files organized by status
• .windsurfrules — Windsurf agent behavior rules
• CLAUDE.md — Claude Code agent behavior rules
• .clinerules — Cline agent behavior rules

Repository Structure

├── Features/              # Kanban board (authoritative source)
├── scripts/               # Onboarding and sync utilities
├── Docs/                  # Additional documentation
├── KANBAN.md             # Main workflow rules for ai agents
└── tasks.csv             # Generated Kanban export

License

MIT — Use this template for your own projects.