{"id":"6de416e8-feb4-402d-a46f-8bae18739f9b","slug":"alesha-pro-quorum","kind":"skill","name":"quorum","summary":"","owner":"alesha-pro","repoUrl":"https://github.com/alesha-pro/quorum","license":"MIT","licenseTier":"permissive","trust":"community","stars":9,"match":null,"fewStars":false,"format":"skill","createdAt":"2026-06-18T08:36:06.000Z","addedAt":"2026-10-21T23:18:04.237Z","logoUrl":"https://avatars.githubusercontent.com/u/68546545?u=9aea5e26580e6ad31ed57ffb91ec350cc21c0dc8&v=4","linkOnly":false,"collection":null,"skillPath":null,"categories":["writing"],"tags":[],"runtime":"binary","local":true,"cloudCapable":false,"cloudCapability":"unknown","cloudReadyOnPlatform":false,"needsApiKey":false,"reviewed":false,"syncedAt":"2026-10-21T23:22:28.829Z","sample":false,"url":"https://conduits.dev/tools/alesha-pro-quorum","language":"Shell","transports":["stdio"],"install":{},"requirements":{"env":[],"upstreamMayCharge":false},"tools":null,"readme":"<p align=\"center\">\n  <img src=\"assets/quorum-banner.png\" alt=\"Quorum — convene a quorum of AI models from your terminal\" width=\"100%\">\n</p>\n\n# Quorum\n\n**Convene a quorum of AI models from your terminal.** A set of portable [Agent Skills](https://opencode.ai/docs/skills/) that let any coding agent summon *other* AI CLIs — OpenAI **Codex**, **opencode** (DeepSeek / Kimi / Qwen), and a fresh **Claude** — as an independent panel for review, research, and hard calls. Plus **`turbo-review`**: one command that runs them all at once and synthesizes the verdicts.\n\n> Where the models agree, you have high confidence. Where they disagree, you get a precise list of things to look at.\n\n## Why\n\nA single model — even a great one — has blind spots and talks itself into things. A second opinion from a *different* model, in a *clean context*, catches what the in-session model glossed over. Quorum makes that one keystroke, and `turbo-review` turns it into a panel vote.\n\n## The skills\n\n| Skill | What it does | Backing CLI |\n|---|---|---|\n| `codex` | Independent second opinion / review from GPT‑5.5 | `codex exec` |\n| `opencode` | External review or research pass | `opencode run` (DeepSeek / Kimi / Qwen) |\n| `claude` | Fresh-context Claude pass — a clean-slate second look | `claude -p` |\n| **`turbo-review`** | **Flagship.** Fans the task out to all available models in parallel, then synthesizes consensus vs. disagreement | all of them |\n| `codex-imagegen` | *(supplementary)* Generate image assets via Codex's native imagegen | `codex exec` |\n\nExternal passes are **read-only** by default: they review, research and propose; your host agent synthesizes and makes the edits. The diff stays coherent, and nothing is touched without you.\n\n## Requirements\n\nEach skill needs its CLI on your `PATH` — install only the ones you want; `turbo-review` uses whichever are present and skips the rest.\n\n- [`codex`](https://developers.openai.com/codex) — for `codex` and `codex-imagegen`\n- [`opencode`](https://opencode.ai) — for `opencode` (free models built in, or bring your own provider)\n- [`claude`](https://claude.com/claude-code) — for `claude`\n\n## Install\n\nThese are standard Agent Skills (`SKILL.md` folders), so they work in **Claude Code**, **Codex**, and **opencode** alike.\n\n### Quick install (all agents)\n\n```bash\ngit clone https://github.com/alesha-pro/quorum\ncd quorum\n./install.sh\n```\n\n`install.sh` symlinks each skill into the shared `~/.agents/skills` directory (read by all three agents) plus the native skills dir of every agent it detects. Restart your agent — or start a new session — afterwards. Undo any time with `./install.sh --uninstall`.\n\n### Manual install\n\nSymlink (or copy) the skill folders into your agent's skills directory:\n\n| Agent | Skills directory |\n|---|---|\n| Claude Code | `~/.claude/skills/` |\n| OpenAI Codex | `~/.agents/skills/` or `~/.codex/skills/` |\n| opencode | `~/.config/opencode/skills/` (also reads `~/.agents/skills/`) |\n\n```bash\nln -s \"$PWD/skills/turbo-review\" ~/.claude/skills/turbo-review\n```\n\n### Claude Code plugin\n\nPrefer the plugin manager? This repo is also a Claude Code marketplace:\n\n```text\n/plugin marketplace add alesha-pro/quorum\n/plugin install quorum@quorum\n```\n\nInstalled via the plugin, the skills are namespaced (`/quorum:turbo-review`, `/quorum:codex`, …). Installed as loose skills, they're bare (`/turbo-review`, `/codex`, …).\n\n## Use\n\n```text\n/turbo-review   \"review my staged diff before I merge — correctness, regressions, prod risk\"\n/codex          \"challenge this migration plan: <plan>\"\n/opencode       \"research the trade-offs of X vs Y for our use case\"\n/claude         \"fresh-eyes pass on this design — what am I missing?\"\n/codex-imagegen \"a 32x32 transparent pixel-art coin sprite → ./assets/coin.png\"\n```\n\nYour agent will also reach for these on its own — running `turbo-review` before a risky change, or asking `codex` to pressure-test a plan — because that's what the skill descriptions tell it to do.\n\n## Configuration\n\n- **opencode model** — defaults to the **free** `opencode/deepseek-v4-flash-free` (opencode Zen, no API key). Run `opencode models` to see what's available to you and pass `--model` for a stronger one you've configured; details in [`skills/opencode/SKILL.md`](skills/opencode/SKILL.md).\n- **codex model** — defaults to your Codex config (e.g. `gpt-5.5`, `xhigh`). Override per call with `-m`.\n- **claude model** — defaults to your Claude Code config; pick a specific one with `--model`.\n\n## Compatibility\n\n| | Claude Code | Codex | opencode |\n|---|:---:|:---:|:---:|\n| Loads `SKILL.md` skills | ✅ | ✅ | ✅ |\n| Quick install via `install.sh` | ✅ | ✅ | ✅ |\n| Plugin marketplace | ✅ | — | — |\n\n## License\n\nMIT © alesha-pro\n"}