Agentic Development

Agentic Development

Spec-driven development orchestration system for Claude Code. Zero-assumption, gate-enforced, professional execution engine.

Overview

Agentic Development is a Claude Code plugin that orchestrates software development through specialized agents, enforcing quality gates and spec alignment at every stage. The system learns from each iteration using persistent memory.

Key Principles:

• Zero assumptions: Gates fail on any ambiguity
• Spec-driven: All code generated from validated specifications
• Atomic commits: One commit per completed task
• Self-learning: Agents track patterns and improve over time
• Token-efficient: Context files prevent token bloat

Installation

Open Claude Code, and type /plugin

> /plugin 
╭─────────────────────────────────────────────────────────────────────────────────────╮
│ Plugins                                                                             │
│                                                                                     │
│   1. Browse and install plugins                                                     │
│   2. Manage and uninstall plugins                                                   │
│ ❯ 3. Add marketplace                                                                │
│   4. Manage marketplaces                                                            │
╰─────────────────────────────────────────────────────────────────────────────────────╯
   Press ↑↓ to navigate · Enter to select · Esc to exit

Navigate to 3. Add marketplace and press Enter.

Type in jsell-rh/agentic-development-plugins and press Enter.

╭─────────────────────────────────────────────────────────────────────────────────────╮
│ Add Marketplace                                                                     │
│                                                                                     │
│ Enter marketplace source:                                                           │
│ Examples:                                                                           │
│  • owner/repo (GitHub)                                                              │
│  • git@github.com:owner/repo.git (SSH)                                              │
│  • https://example.com/marketplace.json                                             │
│  • ./path/to/marketplace                                                            │
│                                                                                     │
│ jsell-rh/agentic-development-plugins                                                │
╰─────────────────────────────────────────────────────────────────────────────────────╯
   Enter to add · Esc to cancel

Follow the rest of the prompts and then restart Claude Code.

Usage

Invoke the orchestrator skill:

Please use the Orchestrator skill to perform development against <my-requirements.md>.

The orchestrator will guide you through the development cycle, asking clarifying questions at gates.

Development Flow

graph TD
    A[User Requirements] --> B{Gate 0: Requirements Refiner}
    B -->|FAIL| A
    B -->|PASS| C[Stage 1: Repo Setup]
    C --> D[Stage 2: Spec Generation]
    D --> E{Gate 3: Spec Refiner}
    E -->|FAIL| F{Gap Type?}
    F -->|Requirements Issue| A
    F -->|Spec Issue| D
    E -->|PASS| G[Stage 4: Development Manager]
    G --> H[Stage 5: Team Lead]

    H --> I{More Tasks?}
    I -->|Yes| J[Create Task Context]
    J --> K[Documentation Expert<br/>Fetch Latest Docs]
    K --> K2{Gate 6: Doc-Spec Alignment}
    K2 -->|FAIL: Modify Spec| D
    K2 -->|FAIL: Clarify| A
    K2 -->|PASS| L{Task Type?}
    L -->|Python| M[Python Expert]
    L -->|FastAPI| N[FastAPI Expert]
    L -->|Deployment| O[Deployment Expert]
    L -->|Security| P[Security Expert]
    L -->|Documentation| Q[Documentation Writer]

    M --> R{Gate 7: Spec Alignment}
    N --> R
    O --> R
    P --> R
    Q --> R

    R -->|FAIL| L
    R -->|PASS| S[Team Lead: Atomic Commit]
    S --> T[Check Off Task]
    T --> I

    I -->|No| U[Project Complete]

    %% Gates in red
    style B fill:#f96,stroke:#333,stroke-width:2px
    style E fill:#f96,stroke:#333,stroke-width:2px
    style K2 fill:#f96,stroke:#333,stroke-width:2px
    style R fill:#f96,stroke:#333,stroke-width:2px

    %% Agents in blue
    style C fill:#69f,stroke:#333,stroke-width:2px
    style D fill:#69f,stroke:#333,stroke-width:2px
    style G fill:#69f,stroke:#333,stroke-width:2px
    style H fill:#69f,stroke:#333,stroke-width:2px
    style K fill:#69f,stroke:#333,stroke-width:2px
    style M fill:#69f,stroke:#333,stroke-width:2px
    style N fill:#69f,stroke:#333,stroke-width:2px
    style O fill:#69f,stroke:#333,stroke-width:2px
    style P fill:#69f,stroke:#333,stroke-width:2px
    style Q fill:#69f,stroke:#333,stroke-width:2px
    style S fill:#69f,stroke:#333,stroke-width:2px

    %% Completion in green
    style U fill:#9f6,stroke:#333,stroke-width:2px

Stages & Agents

Gates (Quality Control)

• Gate 0: requirements-refiner - Validates requirements for completeness
• Gate 3: spec-kit-refiner - Validates specs are generative
• Gate 6: documentation-spec-alignment - Validates specs against latest library docs, flags discrepancies
• Gate 7: spec-alignment-reviewer - Validates code matches specs

Process Agents

• Stage 1: repo-setup-expert - Initializes repo structure, git hooks
• Stage 2: spec-kit-expert - Generates spec files (*.spec.md) from requirements
• Stage 4: development-manager - Creates phased roadmap
• Stage 5: team-lead - Assigns tasks, creates atomic commits

Stage 6 Agents

• documentation-expert (Pre-processor) - Fetches latest docs from internet
• python-expert - Python code + tests, type hints, Pydantic
• fastapi-expert - FastAPI apps using subrouter pattern
• deployment-expert - Dockerfiles (UBI9), Konflux/Tekton CI/CD
• security-expert - Security reviews, OWASP compliance
• documentation-writer - User-facing docs, README, guides

Directory Structure (Generated Projects)

your-project/
├── specs/                    # Spec files (*.spec.md, spec-kit methodology)
├── docs/                     # Documentation
│   ├── roadmap.md           # Phased implementation plan
│   ├── architecture.md      # System architecture
│   └── security.md          # Security documentation
├── src/                      # Source code
├── tests/                    # Test files
├── .agent-context/          # Task context files (ephemeral)
└── .agent-memory/           # Agent learning files (persistent)
    ├── requirements-refiner.md
    ├── spec-kit-expert.md
    └── ...

Agent Memory System

Each agent maintains a memory file (.agent-memory/<agent-name>.md) tracking:

• Patterns observed
• Effective solutions
• Common mistakes
• Project-specific standards

Format:

### 2025-01-12T14:30:00Z
- **Pattern**: What was observed
- **Action**: What to do differently
- **Context**: Where/when this occurred

Agents read their memory at start and append learnings at end. Max 50 entries.

Configuration

Available Stage 6 Agents

The orchestrator knows about these specialists:

• python-expert
• fastapi-expert
• deployment-expert
• security-expert
• documentation-writer

To add more, see AGENTS.md.

Tech Stack Defaults

• Base Images: Red Hat UBI9 (hardened)
• CI/CD: Konflux (Tekton-based)
• Python: 3.13+, Pydantic, Pytest
• FastAPI: Subrouter pattern, dependency injection
• Hooks: rh-hooks-ai + conventional commits

Requirements

• Claude Code CLI
• Git
• (Optional) pre-commit

Contributing

See AGENTS.md for instructions on adding/modifying agents.

License

Apache-2.0