setup
Walks a non-technical user through connecting every judge model this plugin can use — Codex, OpenCode, OpenRouter, Gemini — one at a time, in plain words. Use for "multi setup", "set up multi", "configure judges/reviewers/ backends", "onboarding", or whenever a probe shows missing backends and the user needs to know what to do about it.
szarkans · 已有 AI 审核记录
安装与使用
运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。
npx skills add szarkans/multi --skill setup技能说明
name: setup description: >- Walks a non-technical user through connecting every judge model this plugin can use — Codex, OpenCode, OpenRouter, Gemini — one at a time, in plain words. Use for "multi setup", "set up multi", "configure judges/reviewers/ backends", "onboarding", or whenever a probe shows missing backends and the user needs to know what to do about it. allowed-tools: Bash, Read, Grep, Glob argument-hint: "[nothing needed — just run it]"
Connect the judges
!sh -c 'for p in "$CLAUDE_PLUGIN_ROOT/scripts" "$HOME/.claude/skills/multi/scripts" "./.claude/skills/multi/scripts"; do [ -x "$p/probe.sh" ] && { "$p/probe.sh"; echo "scripts-dir: $p"; exit 0; }; done; echo "probe: NOT FOUND — locate scripts/probe.sh in this plugin and run it yourself"'
$SCRIPTS is whatever the probe printed as scripts-dir:. The probe already
ran above — read it, do not run it again.
The user here is not assumed technical. No jargon, no walls of text, one step at a time, and never make them feel behind for not having done this already.
What this plugin actually is, in plain words
Say this first, briefly, before touching any config:
This plugin asks several different AI models to look at the same thing instead of just one. When they agree, that is a real signal. When they disagree, that disagreement is the whole point — one model alone can't tell you it might be wrong, three models can.
The lineup:
- Claude — already here, always available, nothing to configure.
- GPT, through the Codex CLI — needs a ChatGPT account.
- DeepSeek, through OpenCode — free, works out of the box once installed.
- Gemini, through its own CLI — needs a free API key.
- OpenRouter — the bonus one. A single key unlocks many other model
families at once. Free models by default, or any specific model with
--backend name:model(e.g.--backend openrouter:z-ai/glm-5.2:free).
Fix one thing at a time
Look at the probe output above. For each backend that is missing or not
configured, offer the fix — one at a time, not all four in a list. Start
with the most valuable one that's missing. Codex comes first if it's missing:
/multi:code-review refuses to run at all without it, the others just get
weaker.
Order to work through, skipping each install or login fix already OK. The
OpenCode model-choice check below always runs, even when OpenCode is OK:
-
Codex missing or not logged in
codex loginNeeds a ChatGPT account. This opens a browser to sign in — nothing to type here.
-
OpenCode If it is missing:
curl -fsSL https://opencode.ai/install | bash(or
npm install -g opencode-aiif they'd rather use npm). Nothing to pay for or sign up to — it works with free models right out of the box.Once OpenCode is checked — whether it was already
OKor was just installed — check its model choice. The config is${XDG_CONFIG_HOME:-$HOME/.config}/multi/models, unlessMULTI_MODELS_CONFIGoverrides it. It is a hand-edited file ofkey: value1 value2 ...lines with#comments. Today the only key isopencode:; its first model is primary and the rest are fallbacks.Determine the effective primary model in this exact order: use
MULTI_OPENCODE_MODELwhen it is set; otherwise use the first model on the config'sopencode:line when present; otherwise use the auto-detected primary reported by the probe above. If the config has anopencode:line, still say which models are pinned, in order, but do not use the line's mere presence or absence to decide whether the active channel is free.Decide the warning from the effective primary model's provider prefix:
opencode/*is the free channel andopencode-go/*is the paid Go subscription channel. Warn only when the effective primary is anopencode/*free model. Thus an environment override takes precedence over a configured model, and a configured free primary still gets the warning. For a free effective primary, say this loudly and plainly:Important: you're using free models. They are weaker and less stable. A review may come back thin or even empty, so reviewing on free is a real quality compromise.
Then ask an explicit yes/no question:
Want me to look up the current models, their prices, and how much usage you'd get (how much you'd pay) — and help you pick a good one + fallbacks?
If no, acknowledge that keeping free is a deliberate choice and move on. If yes:
- Run both
opencode --pure modelsandopencode --pure models --verbose;--pureis required so plugin noise does not pollute the output, and the verbose listing includes costs. - Use web search for unfamiliar model names: your own model knowledge is stale and newer models will not be in it. Check what they actually are, their quality and benchmarks, and current pricing.
- Propose one primary model and an ordered fallback chain.
- After the user chooses, create the config directory if needed and write
opencode: <primary> <fallback1> <fallback2> ..., preserving every other line already in the file.
Remind them that this is deliberately hand-editable: they can change that one
opencode:line themselves at any time, with no agent needed. - Run both
-
OpenRouter not configured Key comes from openrouter.ai — sign up, create a key. There's a free tier, no card needed to get one key working. Then they run this themselves, in their own terminal — it asks for the key and does not echo it:
$SCRIPTS/setup.sh set OPENROUTER_API_KEY -
Gemini not configured Key is free, from aistudio.google.com. The second line is theirs to run in their own terminal, same as above:
npm i -g @google/gemini-cli $SCRIPTS/setup.sh set GEMINI_API_KEY
Give one
阅读完整来源 ↗