English | 한국어 | 日本語 | 中文 | Deutsch | Français
Looking for Claude Code? → my-claude — same Boss orchestration in native Claude
.mdagent format
All-in-one agent harness for OpenAI Codex CLI. Install once, 330+ agents ready.
Boss auto-discovers every agent and skill at runtime,
then routes your task to the right specialist via spawn_agent. No config. No boilerplate.
git clone --depth 1 https://github.com/sehoon787/my-codex.git /tmp/my-codex
bash /tmp/my-codex/install.sh
rm -rf /tmp/my-codexcurl -fsSL https://raw.githubusercontent.com/sehoon787/my-codex/main/install.sh | bashBoss is the meta-orchestrator at the core of my-codex. It never writes code — it discovers, classifies, matches, delegates, and verifies.
User Request
│
▼
┌─────────────────────────────────────────────┐
│ Phase 0 · DISCOVERY │
│ Scan ~/.codex/agents/*.toml at runtime │
│ → Build live capability registry │
└──────────────────────┬──────────────────────┘
▼
┌─────────────────────────────────────────────┐
│ Phase 1 · INTENT GATE │
│ Classify: trivial | build | refactor | │
│ mid-sized | architecture | research | ... │
│ → Counter-propose skill if better fit │
└──────────────────────┬──────────────────────┘
▼
┌─────────────────────────────────────────────┐
│ Phase 2 · CAPABILITY MATCHING │
│ P1: Exact skill match │
│ P2: Specialist agent via spawn_agent │
│ P3: Multi-agent orchestration │
│ P4: General-purpose fallback │
└──────────────────────┬──────────────────────┘
▼
┌─────────────────────────────────────────────┐
│ Phase 3 · DELEGATION │
│ spawn_agent with structured instructions │
│ TASK / OUTCOME / TOOLS / DO / DON'T / CTX │
└──────────────────────┬──────────────────────┘
▼
┌─────────────────────────────────────────────┐
│ Phase 4 · VERIFICATION │
│ Read changed files independently │
│ Run tests, lint, build │
│ Cross-reference with original intent │
│ → Retry up to 3× on failure │
└─────────────────────────────────────────────┘
Boss cascades every request through a priority chain until the best match is found:
| Priority | Match Type | When | Example |
|---|---|---|---|
| P1 | Skill match | Task maps to a self-contained skill | "merge PDFs" → pdf skill |
| P2 | Specialist agent | Domain-specific agent exists | "security audit" → security-reviewer |
| P3a | Boss direct | 2–4 independent agents | "fix 3 bugs" → parallel spawn |
| P3b | Sub-orchestrator | Complex multi-step workflow | "refactor + test" → Sisyphus |
| P4 | Fallback | No specialist matches | "explain this" → general agent |
| Complexity | Model | Used For |
|---|---|---|
| Deep analysis, architecture | o3 (high reasoning) | Boss, Oracle, Sisyphus, Atlas |
| Standard implementation | o3 (medium) | executor, debugger, security-reviewer |
| Quick lookup, exploration | o4-mini (low) | explore, simple advisory |
For end-to-end feature implementation, Boss orchestrates a structured sprint:
Phase 1: DESIGN Phase 2: EXECUTE Phase 3: REVIEW
(interactive) (autonomous) (interactive)
───────────────────── ───────────────────── ─────────────────────
User decides scope executor runs tasks Compare vs design doc
Engineering review Auto code review Present comparison table
Confirm "design done" Architect verification User: approve / improve
┌─────────────────────────────────────────────────────┐
│ User Request │
└───────────────────────┬─────────────────────────────┘
▼
┌─────────────────────────────────────────────────────┐
│ Boss · Meta-Orchestrator (o3 high) │
│ Discovery → Classification → Matching → Delegation │
└──┬──────────┬──────────┬──────────┬─────────────────┘
│ │ │ │
▼ ▼ ▼ ▼
┌──────┐ ┌────────┐ ┌────────┐ ┌────────┐
│ P3a │ │ P3b │ │ P1/P2 │ │Config │
│Direct│ │Sub-orch│ │ Skill/ │ │Control │
│2-4 │ │Sisyphus│ │ Agent │ │config. │
│spawn │ │Atlas │ │ Direct │ │toml │
└──────┘ └────────┘ └────────┘ └────────┘
┌─────────────────────────────────────────────────────┐
│ Agent Layer (330+ installed TOML files) │
│ OMO 9 · OMX 33 · Awesome Core 54 · Superpowers 1 │
│ + 20 domain agent-packs (on-demand) │
├─────────────────────────────────────────────────────┤
│ Skills Layer (200+ from ECC + gstack + OMX + more) │
│ tdd-workflow · security-review · autopilot │
│ pdf · docx · pptx · xlsx · team │
├─────────────────────────────────────────────────────┤
│ MCP Layer │
│ Context7 · Exa · grep.app │
└─────────────────────────────────────────────────────┘
| Category | Count | Source |
|---|---|---|
| Core agents (always loaded) | 98 | Boss 1 + OMO 9 + OMX 33 + Awesome Core 54 + Superpowers 1 |
| Agent packs (on-demand) | 220+ | 20 domain categories from agency-agents + awesome-codex-subagents |
| Skills | 200+ | ECC 180+ · gstack 40 · OMX 36 · Superpowers 14 · Core 1 |
| MCP Servers | 3 | Context7, Exa, grep.app |
| config.toml | 1 | my-codex |
| AGENTS.md | 1 | my-codex |
Core Agent — Boss meta-orchestrator (1)
| Agent | Model | Role | Source |
|---|---|---|---|
| Boss | o3 high | Dynamic runtime discovery → capability matching → optimal routing. Never writes code. | my-codex |
OMO Agents — Sub-orchestrators and specialists (9)
| Agent | Model | Role | Source |
|---|---|---|---|
| Sisyphus | o3 high | Intent classification → specialist delegation → verification | oh-my-openagent |
| Hephaestus | o3 high | Autonomous explore → plan → execute → verify | oh-my-openagent |
| Atlas | o3 high | Task decomposition + 4-stage QA verification | oh-my-openagent |
| Oracle | o3 high | Strategic technical consulting (read-only) | oh-my-openagent |
| Metis | o3 high | Intent analysis, ambiguity detection | oh-my-openagent |
| Momus | o3 high | Plan feasibility review | oh-my-openagent |
| Prometheus | o3 high | Interview-based detailed planning | oh-my-openagent |
| Librarian | o3 medium | Open-source documentation search via MCP | oh-my-openagent |
| Multimodal-Looker | o3 medium | Image/screenshot/diagram analysis | oh-my-openagent |
OMC Agents — Specialist workers (19)
| Agent | Role | Source |
|---|---|---|
| analyst | Pre-analysis before planning | oh-my-claudecode |
| architect | System design and architecture | oh-my-claudecode |
| code-reviewer | Focused code review | oh-my-claudecode |
| code-simplifier | Code simplification and cleanup | oh-my-claudecode |
| critic | Critical analysis, alternative proposals | oh-my-claudecode |
| debugger | Focused debugging | oh-my-claudecode |
| designer | UI/UX design guidance | oh-my-claudecode |
| document-specialist | Documentation writing | oh-my-claudecode |
| executor | Task execution | oh-my-claudecode |
| explore | Codebase exploration | oh-my-claudecode |
| git-master | Git workflow management | oh-my-claudecode |
| planner | Rapid planning | oh-my-claudecode |
| qa-tester | Quality assurance testing | oh-my-claudecode |
| scientist | Research and experimentation | oh-my-claudecode |
| security-reviewer | Security review | oh-my-claudecode |
| test-engineer | Test writing and maintenance | oh-my-claudecode |
| tracer | Execution tracing and analysis | oh-my-claudecode |
| verifier | Final verification | oh-my-claudecode |
| writer | Content and documentation | oh-my-claudecode |
Awesome Core Agents (54) — From awesome-codex-subagents
4 categories installed to ~/.codex/agents/:
01-core-development (12) accessibility-tester, ad-security-reviewer, agent-installer, api-designer, code-documenter, code-reviewer, dependency-manager, full-stack-developer, monorepo-specialist, performance-optimizer, refactoring-specialist, tech-debt-analyzer
03-infrastructure (16) azure-infra-engineer, cloud-architect, container-orchestrator, database-architect, disaster-recovery-planner, edge-computing-specialist, infrastructure-as-code, kubernetes-operator, load-balancer-specialist, message-queue-designer, microservices-architect, monitoring-specialist, network-engineer, serverless-architect, service-mesh-designer, storage-architect
04-quality-security (16) api-security-tester, chaos-engineer, compliance-auditor, contract-tester, data-privacy-officer, e2e-test-architect, incident-responder, load-tester, mutation-tester, penetration-tester, regression-tester, security-scanner, soc-analyst, static-analyzer, threat-modeler, vulnerability-assessor
09-meta-orchestration (10) agent-organizer, capability-assessor, conflict-resolver, context-manager, execution-planner, multi-agent-coordinator, priority-manager, resource-allocator, task-decomposer, workflow-orchestrator
Superpowers Agent (1) — From obra/superpowers
| Agent | Role | Source |
|---|---|---|
| superpowers-code-reviewer | Comprehensive code review with brainstorming and TDD verification | superpowers |
Agent Packs — On-demand domain specialists (21 categories)
Installed to ~/.codex/agent-packs/. Managed via:
# View current state
~/.codex/bin/my-codex-packs status
# Enable a pack immediately
~/.codex/bin/my-codex-packs enable marketing
# Switch profiles at install time
bash /tmp/my-codex/install.sh --profile minimal
bash /tmp/my-codex/install.sh --profile full| Pack | Count | Examples |
|---|---|---|
| engineering | 32 | Backend, Frontend, Mobile, DevOps, AI, Data |
| marketing | 27 | Douyin, Xiaohongshu, WeChat OA, TikTok, SEO |
| language-specialists | 27 | Python, Go, Rust, Swift, Kotlin, Java |
| specialized | 31 | Legal, Finance, Healthcare, Workflow |
| game-development | 20 | Unity, Unreal, Godot, Roblox, Blender |
| infrastructure | 19 | Cloud, K8s, Terraform, Docker, SRE |
| developer-experience | 13 | MCP Builder, LSP, Terminal, Rapid Prototyper |
| data-ai | 13 | Data Engineer, ML, Database, ClickHouse |
| specialized-domains | 12 | Supply Chain, Logistics, E-Commerce |
| design | 11 | Brand, UI, UX, Visual Storytelling |
| business-product | 11 | Product Manager, Growth, Analytics |
| testing | 11 | API, Accessibility, Performance, E2E, QA |
| sales | 8 | Deal strategy, pipeline, outbound |
| paid-media | 7 | Google Ads, Meta Ads, Programmatic |
| research-analysis | 7 | Trend, Market, Competitive Analysis |
| project-management | 6 | Agile, Jira, workflows |
| spatial-computing | 6 | XR, WebXR, AR/VR, visionOS |
| support | 6 | Customer support, developer advocacy |
| academic | 5 | Study abroad, corporate training |
| product | 5 | Product management, UX research |
| security | 5 | Penetration testing, compliance, audit |
Skills — 200+ from 5 sources
| Source | Count | Key Skills |
|---|---|---|
| everything-claude-code | 180+ | tdd-workflow, autopilot, security-review, coding-standards |
| oh-my-codex | 36 | plan, team, trace, deep-dive, blueprint, ultrawork |
| gstack | 40 | /qa, /review, /ship, /cso, /investigate, /office-hours |
| superpowers | 14 | brainstorming, systematic-debugging, TDD, parallel-agents |
| my-codex Core | 1 | boss-advanced |
MCP Servers (3)
| Server | Purpose | Cost |
|---|---|---|
| Real-time library documentation | Free | |
 Exa |
Semantic web search | Free 1k req/month |
| GitHub code search | Free |
my-codex includes an Obsidian-compatible knowledge management system. Every project maintains a .briefing/ directory as a persistent memory base.
.briefing/
├── INDEX.md ← Project context, recent decisions
├── sessions/ ← Session summaries (YYYY-MM-DD-topic.md)
├── decisions/ ← Architecture & design decisions
├── learnings/ ← Non-obvious solutions, gotchas
├── references/ ← Web findings, factual data
├── agents/ ← Agent execution logs + daily summaries
└── persona/ ← User Philosophy Vault
├── profile.md ← Agent affinity stats (auto-updated)
├── suggestions.jsonl ← Pending/accepted/rejected suggestions
├── rules/ ← Auto-generated routing preferences
└── skills/ ← Auto-generated persona skills
- Session start — Boss reads
INDEX.mdto load project context - During work — Decisions, learnings, and references are captured as notes
- Session end — Summary written,
INDEX.mdupdated, notes linked with[[wiki-links]] - Persona learning — Usage patterns are detected; after repeated agent use, routing preferences are suggested and can be accepted or rejected via
persona-rule.js
Open your project's .briefing/ folder as an Obsidian vault:
- Open Obsidian → Open folder as vault → select
.briefing/ - Notes appear in the graph view, linked by
[[wiki-links]] - YAML frontmatter provides searchable tags and metadata
- Timeline of decisions and learnings builds automatically over sessions
All notes use YAML frontmatter with date, type, tags, and related fields for structured search.
my-codex bundles content from 8 upstream repositories:
| # | Source | What It Provides |
|---|---|---|
| 1 |  my-claude — sehoon787 |
Sister project. Same Boss orchestration in native Claude .md agent format. Skills, rules, and briefing vault shared across both projects. |
| 2 |  awesome-codex-subagents — VoltAgent |
136 production-grade agents in native TOML format. Already Codex-compatible, no conversion needed. 54 core agents auto-loaded. |
| 3 |  oh-my-codex — Yeachan Heo |
36 skills, hooks, HUD, and team pipelines for Codex CLI. Referenced as architectural inspiration. |
| 4 |  agency-agents — msitarzewski |
180+ business specialist agent personas across 14 categories. Converted from Markdown to native TOML via automated pipeline. |
| 5 |  everything-claude-code — affaan-m |
180+ skills across development workflows. Claude Code-specific content stripped; generic coding skills retained. |
| 6 |  superpowers — Jesse Vincent |
14 skills + 1 agent covering brainstorming, TDD, parallel agents, and code review. |
| 7 |  oh-my-openagent — code-yeongyu |
9 OMO agents (Sisyphus, Atlas, Oracle, etc.). Adapted to Codex-native TOML format. |
| 8 |  gstack — garrytan |
40 skills for code review, QA, security audit, deployment. Includes Playwright browser daemon. |
| Workflow | Trigger | Purpose |
|---|---|---|
| CI | push, PR | Validates TOML agent files, skill existence, and upstream file counts |
| Update Upstream | weekly (Monday) / manual | Runs git submodule update --remote and creates auto-merge PR |
| Auto Tag | push to main | Reads version from config.toml and creates git tag if new |
| Pages | push to main | Deploys docs/index.html to GitHub Pages |
| CLA | PR | Contributor License Agreement check |
| Lint Workflows | push, PR | Validates GitHub Actions workflow YAML syntax |
Features built specifically for this project, beyond what upstream sources provide:
| Feature | Description |
|---|---|
| Boss Meta-Orchestrator | Dynamic capability discovery → intent classification → 4-priority routing → delegation → verification |
| 3-Phase Sprint | Design (interactive) → Execute (autonomous via executor) → Review (interactive vs design doc) |
| Agent Tier Priority | core > omo > omc > awesome-core deduplication. Most specialized agent wins. |
| Cost Optimization | o4-mini for advisory, o3 for implementation — automatic model routing for 330+ agents |
| Agent Telemetry | PostToolUse hook logs agent usage to agent-usage.jsonl |
| Smart Packs | Project-type detection recommends relevant agent packs at session start |
| Agent Pack System | On-demand domain specialist activation via --profile and my-codex-packs helper |
| Codex Attribution | git hooks record Codex-touched files and append AI-Contributed-By: Codex to commit messages |
| CI Dedup Detection | Automated duplicate TOML agent detection across upstream syncs |
git clone --depth 1 https://github.com/sehoon787/my-codex.git /tmp/my-codex
bash /tmp/my-codex/install.sh
rm -rf /tmp/my-codexRe-running the same command refreshes to the latest main build, replaces only my-codex-managed files in ~/.codex/, and removes stale skill copies from ~/.agents/skills/.
On first install, my-codex auto-activates a recommended dev set (engineering, design, testing, marketing, support) and records it in ~/.codex/enabled-agent-packs.txt.
# Minimal profile (core agents only, no packs)
bash /tmp/my-codex/install.sh --profile minimal
# Full profile (all 21 pack categories enabled)
bash /tmp/my-codex/install.sh --profile fullinstall.sh installs a codex wrapper plus global git hooks in ~/.codex/git-hooks/:
prepare-commit-msg— Records files changed during a real Codex sessioncommit-msg— AppendsGenerated with Codex CLI: https://github.com/openai/codexwhen staged files intersect the recorded change setpost-commit— AddsAI-Contributed-By: Codextrailer to qualifying commits
Opt-in Co-authored-by trailer: set both git config --global my-codex.codexContributorName '<label>' and my-codex.codexContributorEmail '<github-linked-email>'. Disable entirely: git config --global my-codex.codexAttribution false. my-codex does not change git user.name, git user.email, or commit author identity.
Every agent is a native TOML file in ~/.codex/agents/:
name = "debugger"
description = "Focused debugging specialist — traces failures to root cause"
model = "o3"
model_reasoning_effort = "medium"
[developer_instructions]
content = """
You are a debugging specialist. Analyze failures systematically:
1. Reproduce the issue
2. Isolate the root cause
3. Propose a minimal fix
4. Verify the fix does not break adjacent behavior
"""Global Codex settings in ~/.codex/config.toml:
[agents]
max_threads = 8
max_depth = 1max_threads— Maximum concurrent sub-agentsmax_depth— Maximum nesting depth for agent-spawns-agent chains
Upstream sources managed as git submodules. Pinned commits tracked in .gitmodules.
| Source | Sync |
|---|---|
| agency-agents | submodule |
| everything-claude-code | submodule |
| oh-my-codex | submodule |
| awesome-codex-subagents | submodule |
| gstack | submodule |
| superpowers | submodule |
How is my-codex different from my-claude?
my-codex and my-claude share the same Boss orchestration architecture and upstream skill sources. The key difference is the runtime: my-codex targets OpenAI Codex CLI with native .toml agent format and spawn_agent delegation, while my-claude targets Claude Code with .md agent format and the Agent tool.
Can I use both my-codex and my-claude?
Yes. They install to separate directories (~/.codex/ and ~/.claude/) and do not conflict. Skills from shared upstream sources are adapted for each platform.
How do agent packs work?
Agent packs are domain-specific agent collections installed to ~/.codex/agent-packs/. On first install, a dev profile is auto-activated. Use my-codex-packs enable <pack> to activate additional packs, or reinstall with --profile full to enable all 21 categories.
How does upstream sync work?
A GitHub Actions workflow runs every Monday, pulling the latest commits from all upstream submodules and creating an auto-merge PR. You can also trigger it manually from the Actions tab.
What models does my-codex use?
Boss and sub-orchestrators (Sisyphus, Atlas, Oracle) use o3 with high reasoning effort. Standard workers use o3 with medium reasoning. Lightweight advisory agents use o4-mini.
If a tool reports invalid SKILL.md files under ~/.agents/skills/, the most common cause is a stale local copy or stale symlink target from an older install.
Remove the affected directories from ~/.agents/skills/ and matching entries under ~/.claude/skills/, then reinstall:
npx skills add sehoon787/my-codex -y -gIf you use the full Codex bundle, rerun install.sh once as well. The full installer refreshes ~/.codex/skills/ and removes stale my-codex-managed copies under ~/.agents/skills/.
Issues and PRs are welcome. When adding a new agent, add a .toml file to codex-agents/core/ or codex-agents/omo/ and update the agent list in SETUP.md. See CONTRIBUTING.md for PR validation steps and Codex commit attribution behavior.
Built on the work of: my-claude (sehoon787), awesome-codex-subagents (VoltAgent), oh-my-codex (Yeachan Heo), agency-agents (msitarzewski), everything-claude-code (affaan-m), oh-my-openagent (code-yeongyu), gstack (garrytan), superpowers (Jesse Vincent), openai/skills (OpenAI).
MIT License. See the LICENSE file for details.