Token Economy Delegation — Development skill for Claude Code
One router for local Ollama, Codex and Claude: bulky jobs go to the cheapest worker whose answer can be checked - verified against the source, escalated on failure.
How to install Token Economy Delegation
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open mixocreative/token-economy-delegation and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Token Economy Delegation does
One router for local Ollama, Codex and Claude: bulky jobs go to the cheapest worker whose answer can be checked - verified against the source, escalated on failure. A skill for Claude Code and Codex.
Alternatives in Development
- Notebook Intelligence — A JupyterLab extension supporting Claude Code, Copilot, Ollama, and OpenAI-compatible LLMs, with MCP, skills 331 ★
- Prun — Run prun: parallel delegation fan-out (Codex-primary, Opus coordinates) 239 ★
- Jarvis Assistant Vocal — Assistant vocal local en francais : Claude ou Ollama (offline), domotique Hue, OBS, agenda, navigateur, appels 189 ★
README
token-economy-delegation
One router for local Ollama, Codex and Claude · a skill for Claude Code and OpenAI Codex
**Your best model shouldn't read 2,000 lines to return twelve.**
**Send every bulky job to the cheapest worker whose answer can be *checked* — and escalate the moment a check fails.**
**最貴的模型不該讀兩千行只為了回傳十二行。大量的機械工作交給「答案可以被驗證」的最便宜工人,驗證失敗就立刻往上交辦。**
[](LICENSE) [](#install) [](#install) [](#what-it-routes) [](#quality) [](#install) [](https://github.com/mixocreative/token-economy-delegation/stargazers)
[Quick start](#quick-start) · [How it works](#how-it-works) · [What it routes](#what-it-routes) · [What it caught](#what-it-caught) · [Quality](#quality) · [FAQ](#faq) · [Install](#install)
claude plugin marketplace add mixocreative/token-economy-delegation && claude plugin install token-economy-delegation@mixocreative
Your agent reads a 2,000-line file to find twelve lines. It counts call sites by hand, summarises a log line by line, waits on a seven-hour test suite, then hits the usage limit and starts over. Most of that work needs no judgment at all, but it is all paid for at your most expensive model's price, out of your most limited resource: its context window.
Install this skill and your agent asks one question before each bulky step: ***could a grep have produced this answer?*** If yes, the job goes to the cheapest worker that can do it on **your** machin
Related Skills
Scvd Store Execution Contract
A behavioral contract for autonomous work. Name the success criteria before acting, spend the cheapest evidenc
Stackwich
Installs an operating architecture into Claude Code as five layers — prompt, context, harness, loop, graph: me
Inspeximus
Your agent's most expensive failure is not forgetting. It is confidently remembering the old answer. inspeximu
Agent Smith Plugin
Claude Code plugin — delegate bulky research & first-draft work to Google Gemini, then review and finish with
Goal Resume
Resume a paused or human-escalated goal (resets iteration budget and stall detector)
Recall Escalation
Assess the operator's current frustration level from their last turn — calm/mild/escalated/high/breaking_point
Related Agents
Delegate
Expert LLM delegation specialist that seamlessly connects to external language models including GPT-4, GPT-3.5
Token Economy Agent
Cost-Classification Auditor. Owns whether a proposed recurring mechanism (per-sprint or per-commit) should be
Topic Specialist
Embodies a domain specialist by loading specified skills, then researches how something works using verified p