![\"Loadout](\"./docs/assets/loadout-unified-workflow-v2.webp\")
\nManage skills for 12 coding agents. Hand off and coordinate work between Claude Code and Codex.
\n\n\n \n
\n ![\"CI](\"https://github.com/VirajMishra1/loadout/actions/workflows/ci.yml/badge.svg\"){kind=icon}
\n ![\"npm](\"https://img.shields.io/npm/v/loadout-ai?color=cb3837&logo=npm\")
![\"npm](\"https://img.shields.io/npm/dm/loadout-ai?color=cb3837&label=downloads\")
![\"GitHub](\"https://img.shields.io/github/stars/VirajMishra1/loadout?style=flat&logo=github\")
![\"Node.js](\"https://img.shields.io/badge/Node.js-%3E%3D20-339933?logo=node.js&logoColor=white\"){kind=icon}
![\"MIT](\"https://img.shields.io/badge/License-MIT-blue.svg\"){kind=icon}
\n
\n Quick start ·\n Two agents ·\n Agent skills ·\n Safety ·\n Commands\n
\n\n## Try it in 30 seconds\n\nYou need Node.js 20 or newer and Git.\n\n```bash\nnpm install --global loadout-ai\nloadout setup --mode stable --details\n```\n\nThe second command detects your coding agents and previews the 30-skill Stable\nloadout. It does not change agent files. Review the plan, then apply it:\n\n```bash\nloadout setup --mode stable --yes --approve-risk\nloadout status\n```\n\nLoadout saves a rollback snapshot before applying changes. Run\n`loadout rollback` to restore the previous managed state.\n\n`--approve-risk` acknowledges the instruction-like files reported by the current\nStable preview. Use it only after reading those findings; it does not replace the\npreview or the pinned-source checks.\n\nA later `--yes` invocation recomputes the plan from pinned sources and current agent and filesystem state; it does not persist or prove identity with the earlier preview.\n\n## What Loadout does\n\n| Capability | What you get |\n| -------------- | --------------------------------------------------------------------- |\n| **Discover** | Find skills, tools, and MCP servers worth reviewing |\n| **Curate** | Inspect, screen, and pin sources before trusting them |\n| **Activate** | Install a focused set for this repository across your agents |\n| **Handoff** | Pass durable tasks with bundled context between Claude Code and Codex |\n| **Coordinate** | Share file ownership, contracts, decisions, and acknowledgements |\n\nThe package manager works across supported agents. Handoff and coordination are\ncurrently designed for Claude Code and Codex working in the same repository.\n\n## Use Claude Code and Codex together\n\n### Handoff: pass a task that survives sessions\n\n```bash\nloadout handoff codex \"write unit tests for auth\" --bundle src/auth.ts src/types.ts --verify \"tests pass\" --verify-command npm --verify-args '[\"test\"]'\n```\n\nCodex sees the task when it checks its inbox, works from the attached bounded\ncontext, and marks it done. Claude Code can then read the result. Handoffs live\nin an append-only project log, so restarting either agent does not erase them.\nBundles contain secret-redacted text and are capped at 50 KiB total; still\nreview them before committing and never attach credential files.\n\n### Coordinate: work at the same time without stepping on files\n\n```bash\nloadout coord start --agents claude-code,codex # preview ownership\nloadout coord start --agents claude-code,codex --yes # apply ownership\nloadout coord snapshot codex # inspect shared state\nloadout coord discuss start \"REST or GraphQL?\" --agents claude-code,codex --rounds 2 --max-turns 5\n```\n\nCoordination is structured shared project state, not shared memory or a merged\ncontext window. Events reach an agent at safe turn boundaries or when it checks\nits snapshot; Loadout does not interrupt a turn in progress.\n\n### Let Claude and Codex debate one decision\n\nYou prompt once and Loadout calls both providers in bounded turns—propose,\ncritique, synthesize—then records the decision. Neither agent can inject into\nthe other's in-progress turn; events arrive at safe turn boundaries. Each\nround consumes paid provider turns from your configured Claude and Codex quota.\nThe bounded design discussion reports its turn budget before it starts.\n\nSee the [live coordination guide](./docs/LIVE_COLLABORATION.md) for ownership,\ncontracts, acknowledgements, discussions, the local dashboard, and limitations.\n\n## Try these prompts\n\nInstall Loadout's two first-party skills once so your agent knows the workflow:\n\n```bash\nloadout skills install loadout-handoff --yes\nloadout skills install loadout-curator --yes\n```\n\nThen ask naturally:\n\n> _\"Hand the test writing to Codex.\"_\n>\n> _\"What did Codex leave for me?\"_\n>\n> _\"Which skills should be active for this repo?\"_\n\nThe curator helps choose a focused active set; the handoff skill manages inboxes\nand coordination without making you relay every command.\n\n## Install and choose your agent\n\nFor a reproducible install, pin the current release:\n\n```bash\nnpm install --global loadout-ai@0.9.4\nloadout setup --mode stable --details\nloadout setup --mode stable --yes --approve-risk\nloadout status\n```\n\nStable is the recommended starting point: 30 selected skill directories from\nfour pinned public sources. Loadout auto-detects supported agents; use\n`--agents` when you want to narrow the destination.\n\nIf anything fails, follow the [user test guide](./docs/USER_TEST_GUIDE.md).\nFor profiles, per-agent paths, MCP configuration, and advanced commands, use the\n[full reference](./docs/REFERENCE.md).\n\n## Safety and trust\n\n**Choose -> Inspect -> Preview -> Apply -> Undo**\n\n- Preview is the default for setup, updates, and removal.\n- Every managed apply creates a rollback snapshot first.\n- Catalog sources are pinned and technically inspected, not declared safe or useful by fiat.\n- Static inspection reports scripts, hooks, binaries, domains, credential references, and unsupported components; it is not a security audit.\n- Project recommendations read bounded local metadata. The documented local flow does not upload project source.\n- MCP servers and executable tools stay behind separate preview, permission, and setup steps.\n- Shared manifests store environment-variable or OS-keychain references, not secret values.\n\n\n\n- All catalog records have identified SPDX licenses. See the [recorded license decisions](./docs/UPSTREAM_LICENSE_DECISIONS.md) for the source-by-source record.\n\n\n\nRead the [security policy](./SECURITY.md), [catalog policy](./docs/CATALOG_POLICY.md),\nand [credential and update policy](./docs/CREDENTIAL_AND_UPDATE_POLICY.md) before\ntrusting third-party content.\n\n## Demo\n\n\n \n ![\"Watch](\"https://img.youtube.com/vi/opNqJKX7xMw/maxresdefault.jpg\")
\n \n