{
  "markdown": "<p align=\"center\">\n  <img src=\"assets/cauchy.svg\" alt=\"Cauchy\" width=\"64\">\n</p>\n\n# awow — Agentic Way of Working\n\n<p align=\"center\"><em>Helping humans work better together, using AI.</em></p>\n\nawow gives a coding agent your team's context: the board it reads, the\nconventions it follows, and commands for the work that happens between people.\n\nA coding agent is only as useful as what it starts with, and on most teams that\nstarting context is scattered. The board lives in one tool, the conventions live\nin heads and wiki pages, and every agent product — Claude Code, Codex, Pi,\nCopilot, opencode — wants its own copy of the instructions. Each copy drifts,\nand every session begins from zero.\n\nawow packages the missing pieces and ships them as a plugin:\n\n- **Working rules** every session starts from — go to the board before starting\n  work, write or update the ticket, keep the admin current as you go.\n- **Commands** for the work that happens between people: a board item walked\n  from refinement to an opened PR, a refinement session prepared in advance, a\n  meeting recording turned into decisions, owners, and board items.\n- **A `context/` folder your team owns** — board wiring, mission, conventions,\n  members — that every command reads, and that stays yours rather than being\n  baked into the tool.\n\nIt is markdown throughout, authored in one source tree and built into a package\neach agent product installs, so there is one copy to keep current instead of\nfive. Nothing the agent drafts reaches your board or your repo until you have\nseen it and approved it.\n\nThe full technical guide lives [here](guides/README.md).\n\n## Before you install\n\n- **A supported harness**, installed and signed in: Claude Code, Codex, Pi,\n  the GitHub Copilot CLI, or opencode.\n- **The `gh` CLI, authenticated** (`gh auth login`). Commands use it to open\n  PRs, and it doubles as the board surface for GitHub-hosted boards.\n- **Access to your team's board** — an account that can read and write it.\n  Nothing needs wiring yet: `/setup-awow`'s first required step installs and\n  verifies the surface the agent will use — an MCP server for Linear, Jira,\n  or Azure DevOps, or `gh` for GitHub Issues and Projects — and records it\n  in `context/tooling/board.md`.\n\n## Installing the plugin\n\nA **plugin** is how an agent product installs and versions an extension: one\nbundle of commands, skills, and supporting files, added from a **marketplace**\n(a catalog of plugins) and updated like any other dependency. awow ships as one\nsuch bundle. Each product documents its own plugin model —\n[Claude Code](https://code.claude.com/docs/en/discover-plugins),\n[Codex](https://learn.chatgpt.com/docs/plugins),\n[Pi](https://github.com/earendil-works/pi/blob/main/packages/coding-agent/docs/packages.md)\n(which calls them *packages*), and\n[GitHub Copilot CLI](https://docs.github.com/en/copilot/concepts/agents/copilot-cli/about-cli-plugins).\n\nTo install awow, run the commands for the agent product you use — one block\nonly, not all five:\n\nClaude Code:\n\n```\n/plugin marketplace add CauchyIO/awow\n/plugin install awow@awow\n```\n\nCodex:\n\n```\ncodex plugin marketplace add https://github.com/CauchyIO/awow-dist\ncodex plugin add awow@awow\n```\n\nPi:\n\n```\npi install git:github.com/CauchyIO/awow-dist\n```\n\nGitHub Copilot (requires the Copilot CLI):\n\n```\ncopilot plugin marketplace add CauchyIO/awow\ncopilot plugin install awow@awow\n```\n\nopencode:\n\n```\nopencode plugin awow@git+https://github.com/CauchyIO/awow-dist.git\n```\n\nClaude Code and Copilot install from this repo, which carries the marketplace\nmanifest both read. Codex, Pi and opencode install from `awow-dist`, which\ncarries the built payload. Copilot exposes the commands as skills rather than\nslash commands.\n\n## First: run `/setup-awow`\n\n`/setup-awow` is the first command to run after installing the plugin. It\nwires your board (Linear, Jira, Azure DevOps, GitHub Issues) and writes your\nmission, conventions, and members into `context/` — the context every other\ncommand reads.\n\nIts first question is which of two shapes you want, and it records the answer:\n\n1. **Standalone.** awow set up for one repo, with its own context and board\n   wiring.\n2. **Anchored.** One centralized repo — the **anchor** — holds the shared\n   `context/`, and other repos register as **anchored repos** and read the\n   anchor's context instead of carrying their own. For teams who want a single\n   agentic core across several repositories. The details are in\n   [Setup & the plugin model](guides/guide-setup-and-two-harnesses.md).\n\nRun it once in every repo that uses awow — but it does different work depending\non the repo. In a standalone repo or an **anchor**, it walks the full setup and\nwrites that repo's own `context/`. In an **anchored** repo it detects the anchor\nfrom the root `AGENTS.md` and runs a short registration track instead: it\nrecords which anchor the repo belongs to and which board scope it maps to, then\ntakes the board wiring, conventions and members from the anchor rather than\nbuilding a second copy. So an anchored team sets up the anchor first, then runs\n`/setup-awow` again — briefly — in each repo that anchors to it.\n\nHow you run it is up to you — the format, the pace, and how far you take it:\n\n- Choose a guided walkthrough, or a 25–30 minute team workshop whose\n  transcript becomes the same gated setup proposals.\n- It is incremental and resumable: stop after any step, pick up where you\n  left off.\n- Only Steps 0 and 1 (install shape and board) are required; the rest are\n  recommended in any order.\n\nThe other commands do run without setup — they ask for what's missing and\ncarry on — but they work better with it.\n\n## Then: explore the commands\n\nThe commands work in any repo (anchor or anchored).\n\n| | |\n|---|---|\n| `/my-work` | what the board says needs you, grouped by blocked, waiting, or yours now |\n| `/process-workitem` | a board item from refinement through a planned change to an opened PR |\n| `/refinement-prep` | a brief or a deck broken into right-sized stories before the session |\n| `/process-transcript` | a meeting recording turned into decisions, owners, and board items |\n| `/solution-design-flow` | an architecture argument turned into a decision record |\n| `/artifact` | a deck, one-pager, or report as HTML or PDF |\n\nEach command carries a description of the situation it applies to, so you can\ndescribe what you need instead of typing the command name.\n\nThose six are the ones most teams reach for first; `.agents/commands/` holds\ntwenty-two in all. `/daily-checkin`, `/handover`, `/process-retro`,\n`/board-lifecycle`, `/strategy-flow` and `/okr-cascade` are among the rest. The full set, grouped by\nthe adoption phase each belongs to, is catalogued in\n[`.agents/commands/`](.agents/commands/README.md).\n\n## What the agent picks up\n\n- **awow's [working rules](.agents/skills/using-awow/SKILL.md)**, read at the\n  start of every session: go to the board before starting work, write or update\n  the ticket, and keep the admin current while you work.\n- **Your team context**, where it exists — the `context/` that `/setup-awow`\n  writes, read by whichever commands need it.\n- **An optional OKF catalog**, which routes commands to canonical repositories,\n  SharePoint, or vector-backed sources without copying that material into the\n  anchor.\n\n## Going deeper\n\nThe comprehensive guides live in [`guides/`](guides/README.md) — plain markdown, readable\ndirectly on GitHub or as agent context.\n\nOnce the commands are part of how the team works, the next question is usually\nwhether they are helping. `awow-telemetry` is a second plugin for measuring how\nthe way of working is going: session timelines, prompt-quality review, usage\ncoaching. It runs on Claude Code only.\n\n## Contributing to awow\n\n- **`.agents/` is the source.** `tools/gather.py` builds it into the payloads\n  under `dist/` and `dist-telemetry/`. CI fails on drift with `--check`.\n- **Nothing is mirrored into this repo's `.claude/` or `.github/`.** The\n  marketplace that Claude Code and Copilot install from *is* this repo. A merge\n  to `main` is therefore what reaches a maintainer's own sessions, after\n  `/plugin marketplace update awow` and then `/plugin update awow`.\n- **`/test-awow` is the one exception.** The eval runner lives in this repo's\n  `.claude/commands/` rather than in the payload.\n\nTo exercise a branch's payload before it merges:\n\n```bash\npython tools/gather.py && claude --plugin-dir dist\n```\n\n## Status\n\n**v0.13.0 — pre-1.0, in use.** Working end to end: the installs on all five\nharnesses, the command set, canonical knowledge-source routing, the session\ncontext, and the build with its drift check in CI.\n\nBeing pre-1.0, names and file shapes can still change between releases. The\none migration currently in flight is the `hub` → `anchor` rename: the legacy\nspellings (`hub:`, `$AWOW_HUB`, `.awow/hub.json`, `{HUB}`) are still accepted\neverywhere, so pre-rename repos need no action.\n\nWhat each release changed is in [`CHANGELOG.md`](CHANGELOG.md).\n\n## License\n\nMIT. See [`LICENSE`](LICENSE).\n",
  "bytes": 9051,
  "sha": "451cda7198d3d7460de4d2273046d2488e5c917f710de27694939bf64f156838",
  "repo_slug": "cauchyio/awow",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/okf_cauchyio_awow_context_knowledge_sources__9ba5b978/readme"
}