Skip to content
 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

144 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

agent-os-og

Your system for spec-driven agentic development.

Agent OS transforms AI coding agents from confused interns into productive developers. With structured workflows that capture your standards, your stack, and the unique details of your codebase, Agent OS gives your agents the specs they need to ship quality code on the first try—not the fifth.

Use it with:

  • Claude Code, Cursor, or any other AI coding tool.
  • New products or established codebases.
  • Big features, small fixes, or anything in between.
  • Any language or framework.

Installation

1. Base Installation

Install Agent OS to your home directory:

curl -sSL https://raw.githubusercontent.com/dchuk/agent-os/main/scripts/base-install.sh | bash

This creates ~/agent-os/ with:

  • Configuration in config.yml
  • Profiles in profiles/
  • Installation scripts in scripts/

2. Project Installation

Navigate to your project and install Agent OS:

cd /path/to/your/project
~/agent-os/scripts/project-install.sh

Options:

# Dry run - see what would be installed
~/agent-os/scripts/project-install.sh --dry-run

# With specific options
~/agent-os/scripts/project-install.sh --claude-code-commands true --use-claude-code-subagents true

# See all options
~/agent-os/scripts/project-install.sh --help

3. Update an Existing Project

cd /path/to/your/project
~/agent-os/scripts/project-update.sh

What's New in v3.0.0

Version 3.0.0 introduces structured JSON schemas, bidirectional linking, and a findings system for capturing institutional knowledge.

Breaking Changes

  • roadmap.md replaced with roadmap.json
  • tasks.md replaced with tasks.json
  • New spec-meta.json required in each spec folder
  • New findings.json in product folder

JSON Schemas for Core Data

All primary data files now use JSON with defined schemas:

File Location Purpose
roadmap.json agent-os/product/ Product feature roadmap with status tracking
tasks.json agent-os/specs/[spec]/ Task breakdown for each spec
findings.json agent-os/product/ Institutional knowledge captured during implementation
spec-meta.json agent-os/specs/[spec]/ Spec metadata and roadmap linkage

Bidirectional Linking

Specs and roadmap items are now explicitly linked:

roadmap.json                          spec-meta.json
┌─────────────────────┐              ┌─────────────────────┐
│ id: "roadmap-003"   │◄────────────►│ roadmapItemId:      │
│ specPath: "agent-os │              │   "roadmap-003"     │
│   /specs/2025-01-10 │              │ specId: "2025-01-10 │
│   -project-org"     │              │   -project-org"     │
└─────────────────────┘              └─────────────────────┘

Findings System

Agents now capture institutional knowledge during implementation:

  • Build configuration gotchas
  • Error patterns and solutions
  • Established code patterns
  • Testing strategies
  • Architecture decisions

Findings are:

  • Automatically captured post-implementation
  • Deduplicated and merged when similar
  • Reviewed periodically via /review-findings
  • Synced to AGENTS.md for agent context

File Structure

agent-os/
├── product/
│   ├── mission.md           # Product mission
│   ├── roadmap.json         # JSON roadmap with status tracking
│   ├── findings.json        # Captured learnings
│   └── tech-stack.md        # Tech stack
├── specs/
│   └── [YYYY-MM-DD-spec-name]/
│       ├── spec-meta.json   # Spec metadata and roadmap link
│       ├── spec.md          # Specification document
│       ├── tasks.json       # JSON task breakdown
│       ├── planning/
│       │   ├── requirements.md
│       │   └── visuals/
│       ├── implementation/
│       └── verification/
├── schemas/                  # JSON Schema definitions
│   ├── roadmap.schema.json
│   ├── tasks.schema.json
│   ├── findings.schema.json
│   └── spec-meta.schema.json
└── standards/               # Your coding standards

Schema Reference

roadmap.json

{
  "schemaVersion": "1.0.0",
  "productName": "MyApp",
  "lastUpdated": "2025-01-11T14:30:00Z",
  "items": [
    {
      "id": "roadmap-001",
      "title": "User Authentication",
      "description": "Email/password auth with session management",
      "status": "completed",
      "effort": "M",
      "priority": 1,
      "dependencies": [],
      "specPath": "agent-os/specs/2025-01-05-user-auth",
      "tags": ["backend", "security"],
      "createdAt": "2025-01-02T10:00:00Z",
      "completedAt": "2025-01-08T16:30:00Z"
    }
  ]
}

Status values: plannedspeccedin-progresscompleted | deferred

Effort scale: XS (1 day), S (2-3 days), M (1 week), L (2 weeks), XL (3+ weeks)

tasks.json

{
  "schemaVersion": "1.0.0",
  "specId": "2025-01-10-project-org",
  "specTitle": "Project Organization",
  "roadmapItemId": "roadmap-003",
  "status": "in-progress",
  "summary": {
    "totalTaskGroups": 4,
    "completedTaskGroups": 2,
    "totalTasks": 12,
    "completedTasks": 7
  },
  "taskGroups": [
    {
      "id": "tg-001",
      "name": "Database Layer",
      "layer": "database",
      "dependencies": [],
      "status": "completed",
      "tasks": [...]
    }
  ]
}

Layer values: database, api, frontend, testing, infrastructure, integration, other

findings.json

{
  "schemaVersion": "1.0.0",
  "lastUpdated": "2025-01-11T14:00:00Z",
  "lastReviewedAt": null,
  "findings": [
    {
      "id": "finding-001",
      "category": "code-pattern",
      "title": "Form validation uses Zod schemas",
      "description": "The project uses Zod schemas in src/schemas/...",
      "context": "Any new form or API endpoint needing validation",
      "recommendation": "Create schema in src/schemas/[feature].schema.ts",
      "confidence": "high",
      "status": "active"
    }
  ]
}

Categories: build-config, error-pattern, code-pattern, dependency, performance, testing, architecture, tooling, security, other

Confidence: low, medium, high

spec-meta.json

{
  "schemaVersion": "1.0.0",
  "specId": "2025-01-10-project-org",
  "title": "Project Organization",
  "roadmapItemId": "roadmap-003",
  "status": "in-progress",
  "createdAt": "2025-01-10T15:00:00Z",
  "shapedAt": "2025-01-10T15:45:00Z",
  "speccedAt": "2025-01-10T16:30:00Z",
  "taskedAt": "2025-01-10T17:00:00Z",
  "implementationStartedAt": "2025-01-11T09:00:00Z",
  "completedAt": null,
  "findingsGenerated": ["finding-005"]
}

Status values: draftingshapedspeccedtaskedin-progresscompleted | abandoned


Commands

Command Description
/plan-product Create product mission, roadmap, and tech stack
/shape-spec Shape requirements for a feature
/write-spec Write the specification document
/create-tasks Break spec into tasks.json
/implement-tasks Implement all tasks
/orchestrate-tasks Advanced multi-agent task orchestration
/review-findings Review and maintain findings database

Migration from v2.x

If you have existing markdown roadmaps and task files:

  1. Roadmap: Manually convert roadmap.md to roadmap.json format
  2. Tasks: Existing specs can continue with markdown tasks; new specs will use JSON
  3. Findings: Start fresh - findings.json is initialized automatically

The installation scripts will:

  • Install JSON schemas to agent-os/schemas/
  • Initialize empty findings.json if it doesn't exist
  • Not overwrite existing product files

Documentation & Support


Created by Brian Casel @ Builder Methods

Created by Brian Casel, the creator of Builder Methods, where Brian helps professional software developers and teams build with AI.

Get Brian's free resources on building with AI:

Join Builder Methods Pro for official support and connect with our community of AI-first builders.

About

Agent OS is a system for better planning and executing software development tasks with your AI agents.

Resources

Code of conduct

Contributing

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages