Model router

Skill philgoodvibe/aiai-mastermind-tools-and-skills/skills/model-router/skills/model-router

How to run Claude Code sessions, one-shot workers, and subagents on any model through the local proxy - model IDs, env config, and the three usage modesFrom its SKILL.md

Install

npx -y skills add philgoodvibe/aiai-mastermind-tools-and-skills --skill model-router

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

5 things to look at

skips confirmationTells the agent to proceed without asking first, 1 time: "--permission-mode acceptEdits".
reads credentialsReads from 1 credential source: `MODEL_ROUTER_KEY`.
2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
runs commandsInstructs the agent to run 6 commands, including `claude -p "the task" --model ... --effort ... --bare --max-turns 20 --permission-mode acceptEdits` and 5 more.
fetches URLsInstructs the agent to fetch 1 URL, including http://127.0.0.1:8317/v1/models.

SKILL.md

2.8 KB, 734 tokens by cl100k_base, as published. Nobody here has run it

Model Router - routing knowledge

This skill teaches the three ways to run work on a routed model (any model your proxy serves), and how to switch models. The proxy is CLIProxyAPI running locally; it holds the user's OTHER subscription logins (ChatGPT/Codex, Gemini, Grok, Kimi). The main Claude session always stays on its normal direct login.

Configuration (shared by everything in this plugin)

Env var	Default	Meaning
`MODEL_ROUTER_URL`	`http://127.0.0.1:8317`	proxy address
`MODEL_ROUTER_KEY`	`my-proxy-key`	an `api-keys` entry from the proxy's config.yaml
`MODEL_ROUTER_MODEL`	`gpt-5.6-sol`	default routed model
`MODEL_ROUTER_EFFORT`	`low`	default worker effort

Mode 1 - orchestrator + worker (the flagship): `/fabsol <brief>`

Your main model plans and verifies; the routed model does 100% of the building via headless calls. Cheapest way to ship real builds - see the fabsol command.

Mode 2 - one-shot worker run (any model, one task)

ANTHROPIC_BASE_URL="${MODEL_ROUTER_URL:-http://127.0.0.1:8317}" \
ANTHROPIC_AUTH_TOKEN="${MODEL_ROUTER_KEY:-my-proxy-key}" \
claude -p "the task" \
  --model "${MODEL_ROUTER_MODEL:-gpt-5.6-sol}" \
  --effort "${MODEL_ROUTER_EFFORT:-low}" \
  --bare --max-turns 20 --permission-mode acceptEdits

Or use the wrapper: scripts/worker.sh "the task" [model] [effort] (scripts/worker.ps1 on Windows).

Mode 3 - full session on a routed model (terminal or VSCode)

scripts/session.sh [model] (or session.ps1) launches a whole Claude Code terminal session on the routed model. scripts/vscode.sh [folder] (or vscode.ps1) does the same for a VSCode window - it uses a separate VSCode profile dir because an already-running VSCode ignores new env vars. In such a session, the plugin's worker subagent also resolves, so subagents can run on routed models too.

Switching models

Any ID the proxy serves works everywhere a model is named. Discover what your proxy serves:

curl -s "${MODEL_ROUTER_URL:-http://127.0.0.1:8317}/v1/models" \
  -H "Authorization: Bearer ${MODEL_ROUTER_KEY:-my-proxy-key}"

Only logged-in providers appear. Set MODEL_ROUTER_MODEL to switch the default everywhere at once, or pass a model per call. See the README's model table for provider login commands.

Hard rule

NEVER set ANTHROPIC_BASE_URL / ANTHROPIC_AUTH_TOKEN on the main session's environment or route a Claude subscription login through the proxy - Anthropic's terms ban subscription OAuth in third-party tools. Routed env vars belong on individual headless calls (or a deliberate scripts/session launch for non-Claude models) only.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.