{
  "markdown": "# AgentsAtlas\n\n[![npm version](https://img.shields.io/npm/v/agents-atlas)](https://www.npmjs.com/package/agents-atlas)\n[![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](https://opensource.org/licenses/MIT)\n[![Node >= 16](https://img.shields.io/badge/node-%3E%3D16-brightgreen.svg)](https://nodejs.org)\n\n**A minimal workflow for building production-ready projects with Claude Code & Codex.**\n8 commands. Fresh context via subagent execution. Quality execution.\n\n<img width=\"476\" height=\"302\" alt=\"AgentsAtlas\" src=\"https://github.com/user-attachments/assets/c0936781-d5f4-403b-91ce-7d1328945265\" />\n\n## Quick Start\n\n```bash\nnpx agents-atlas --global\n```\n\nThen in any project:\n\n```bash\n/atlas:init       # Set up project with CLAUDE.md + STATE.md\n/atlas:plan       # Create a 3-5 task plan for current phase\n/atlas:execute    # Run plan with fresh subagent context\n```\n\n## What You Get\n\n- **8 slash commands** that manage any project end-to-end\n- **Fresh 200k-token context** per task via subagent execution — no degradation\n- **XML task format** with built-in verification steps\n- **16 development skills** (debugging, testing, security audit, and more)\n- **Works with both Claude Code and Codex**\n\n## Table of Contents\n\n- [Installation](#installation)\n- [Commands](#commands-claude-code)\n- [Using with Codex](#using-with-codex)\n- [Skills](#skills)\n- [Workflow](#workflow)\n- [Project Structure](#project-structure)\n- [Task Format](#task-format)\n- [Why This Works](#why-this-works)\n- [Customization](#customization)\n- [License](#license)\n\n## Installation\n\n### Option 1: npx (Recommended)\n\n```bash\n# Install globally (available in all projects)\nnpx agents-atlas --global\n\n# Or install locally (current project only)\nnpx agents-atlas --local\n\n# Global commands + global Codex skills\nnpx agents-atlas --global --with-codex-global\n```\n\n`--local` also installs Codex wrappers in your repo (`.agents/skills/atlas-*`, `scripts/atlas-*`, and `.codex/config.toml`).\nUse `--with-codex-global` when installing globally to install Atlas skills in `~/.codex/skills/atlas-*` so they are available across projects (project-local skills can still override them).\n\n### Option 2: From GitHub\n\n```bash\ngit clone https://github.com/AvinashP/AgentsAtlas.git\ncd AgentsAtlas\nnode bin/install.js --global\n```\n\n### Option 3: Manual Symlink\n\n```bash\ngit clone https://github.com/AvinashP/AgentsAtlas.git\nln -s $(pwd)/AgentsAtlas/commands/atlas ~/.claude/commands/atlas\n```\n\n## Commands (Claude Code)\n\n| Command | Purpose |\n|---------|---------|\n| `/atlas:init` | Initialize project with CLAUDE.md + STATE.md |\n| `/atlas:plan` | Create executable plan (3-5 tasks max) |\n| `/atlas:execute` | Run plan with fresh subagent context |\n| `/atlas:status` | Check progress, get next action |\n| `/atlas:review` | Review code changes, capture learnings for CLAUDE.md |\n| `/atlas:sync` | Restore context after `/clear` |\n| `/atlas:triage` | Pull and triage issues from Sentry, GitHub, etc. |\n| `/atlas:complete` | Complete milestone, archive, and prepare for next work |\n\n### Non-Interactive Mode\n\nAdd `--auto` when running without user interaction (CI, scripts, `--dangerously-skip-permissions`):\n\n```bash\n/atlas:init --auto    # Detects project from config files, uses [TODO] for unknowns\n/atlas:plan --auto    # Picks best approach automatically, no prompts\n```\n\nWithout `--auto`, these commands ask interactive questions that get phantom answers in non-interactive mode.\n\n## Using with Codex\n\nAtlas appears in Codex as **skills**, not `/atlas:*` slash commands.\n\n1. Open Codex in a project.\n2. Run `/skills` (or press `$`) and select any `atlas-*` skill.\n\nSkill precedence:\n- **Global skills**: `~/.codex/skills/atlas-*` (installed by `--global --with-codex-global`)\n- **Project-local skills**: `.agents/skills/atlas-*` (override global skills when present)\n\nQuick verification:\n\n```bash\nls ~/.codex/skills/atlas-plan/SKILL.md\n```\n\nSee [docs/ai-workflow.md](docs/ai-workflow.md) for the full Codex integration guide.\n\n## Skills\n\nAgentsAtlas includes **16 development skills** that provide disciplined workflows:\n\n| Category | Skills |\n|----------|--------|\n| **Discipline** | `debugging`, `testing`, `verifying` |\n| **Process** | `brainstorming`, `receiving-feedback` |\n| **Git** | `committing`, `creating-pr` |\n| **Code Quality** | `refactoring`, `security-audit` |\n| **Utility** | `explaining-code`, `scaffolding` |\n| **Building** | `frontend-design`, `mcp-builder`, `web-artifacts-builder`, `webapp-testing`, `skill-creator` |\n\nThe first 11 skills emphasize the **Iron Law** approach — finding root causes before fixes, evidence before claims, and test-driven development. The building skills (from Anthropic) provide production-grade UI design, MCP server creation, and testing capabilities.\n\n## Workflow\n\n```\n/atlas:init          # Once per project\n    ↓\n/atlas:plan          # Plan current phase (3-5 tasks)\n    ↓\n/atlas:execute       # Execute with fresh context\n    ↓\n/atlas:status        # What's next?\n    ↓\n(repeat plan → execute → status)\n```\n\n### Bug Triage Workflow\n\n```\n/atlas:triage        # Pull issues from Sentry/GitHub/JIRA\n    ↓\n/atlas:plan          # Plan fixes for selected issues\n    ↓\n/atlas:execute       # Execute fix plan\n```\n\n### Milestone Completion\n\n```\n(all phases complete)\n    ↓\n/atlas:complete      # Archive, tag, prepare for next work\n    ↓\n/atlas:plan          # Start new milestone\n```\n\n### Code Review\n\n```\n/atlas:review        # Review current phase's changes\n    ↓\n(address issues if any)\n    ↓\n/atlas:complete      # Or continue to next phase\n\n# Or review a PR:\n/atlas:review --pr 123  # Review GitHub PR, suggest CLAUDE.md additions\n```\n\n### Context Management\n\n```\nWhen context degrades:\n1. Run /clear\n2. Run /atlas:sync\n3. Continue working\n\nThe subagent execution pattern means each plan runs\nwith fresh 200k tokens—no accumulated noise.\n```\n\n## Project Structure\n\nAfter `/atlas:init`, your project will have:\n\n```\nyour-project/\n├── CLAUDE.md              # Project context (< 30 lines)\n└── .planning/\n    ├── STATE.md           # Current position + recent decisions\n    ├── ROADMAP.md         # Phase checklist\n    └── phases/\n        ├── 01-PLAN.md     # Phase 1 execution plan\n        ├── 02-PLAN.md     # Phase 2 execution plan\n        └── ...\n```\n\n## Task Format\n\nPlans use XML for clarity:\n\n```xml\n<task id=\"1\">\n<name>Create user authentication endpoint</name>\n<files>src/api/auth.ts, src/lib/jwt.ts</files>\n<action>\n1. Create POST /api/auth/login endpoint\n2. Validate email/password against database\n3. Generate JWT token on success\n4. Return token with 24h expiry\n</action>\n<verify>curl -X POST /api/auth/login -d '{\"email\":\"test@example.com\",\"password\":\"test\"}'</verify>\n<done>Login returns 200 + JWT token; invalid credentials return 401</done>\n</task>\n```\n\n- `verify` = How to test (commands to run)\n- `done` = Acceptance criteria (what success looks like)\n\n## Why This Works\n\n- **Small plans (3-5 tasks)** = Focused execution, fewer mistakes\n- **Fresh context per task** = Consistent quality, no degradation across tasks\n- **Verification steps** = Catch issues immediately, not at the end\n- **Atomic commits** = Easy rollback, clear git history\n- **STATE.md** = Session continuity without context bloat\n\nSee [Setup Tips](docs/setup-tips.md) for hooks, permissions, and advanced Claude Code configuration.\n\n## Customization\n\nEdit the command files in `commands/atlas/` to adjust:\n- Task limits\n- Commit format\n- Deviation rules\n- Output format\n\nCodex wrapper metadata source of truth:\n- `commands/atlas/manifest.json`\n- Run `npm run sync:atlas` after manifest changes.\n\nArtifacts regenerated by the sync script:\n- `.agents/skills/atlas-*`\n- `scripts/atlas-*`\n- `AGENTS.md`\n- `docs/ai-workflow.md`\n- `.claude-plugin/plugin.json` command list\n\nMaintainer checks:\n- `npm run check:atlas`\n\n## License\n\nMIT\n",
  "bytes": 7842,
  "sha": "793e860b5cc48ba4586a4b9cd28e80eb8231a10e47944e73cf2b5a4cce029337",
  "repo_slug": "avinashp/agentsatlas",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/plg_avinashp_agentsatlas_agentsatlas_f61af8bc/readme"
}