todo-jev
Experimental task router that uses Jev, skill profiles and environment checks to send requests to a rule, a skill or a large model.
Overview
todo-jev classifies an incoming request and recommends one of three handling tiers: a local rule, a structured decision or specialist skill, or a foundation model. It combines Jev's classification with structured skill profiles that state when a skill applies and when it should be excluded, plus basic preflight checks of the local environment.
It is an experiment for agent builders who have many skills and want a cheap way to choose the next step. It ships 20 bundled skill profiles, discovers locally installed skills, and has a keyword-based offline fallback so the CLI works without a TypeSafe key. The README is clear that only classification, skill matching and routing are implemented; the execution handlers return example responses, and running the chosen task needs your own rule functions, skill runner or model client.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which tier should handle this request? | Tier 1 local rule · Tier 2 structured decision or skill · Tier 3 foundation model |
| Choice | Which skill profile best fits this request? | — |
How it works
- Run python -m app.cli route "<request>".
- Jev (or the offline keyword fallback) classifies the request.
- Skill profiles and environment checks narrow or exclude skills.
- The CLI prints the recommended tier and skill; execution is left to the user.
Strengths
- Offline fallback makes it usable without an API key.
- Skill profiles encode explicit apply and exclude conditions.
- Environment checks add evidence to the routing choice.
Caveats
- Execution handlers are stubs returning example responses.
- Described as an experiment.
- Examples and docs are partly in Korean.
Getting started
git clone https://github.com/maker-KK/todo-jev.git && pip install -r requirements.txt && python -m app.cli route "<request>"At a glance
- Type
- Agent skill
- Status
- prototype
- Language
- Python
- Runtime
- CLI
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
A task-routing experiment combining skill conditions and environment checks to suggest rules, skills or a large model.
logicrw/awesome-jev-projects⚡ Ultra-fast, low-cost intelligent task classifier and 3-tier routing engine powered by TypeSafe Jev (System One).
hellogumbo/awesome-jev
Frequently asked questions
What does todo-jev decide?
Whether a request should go to a local rule, a structured decision or skill, or a foundation model, and which skill fits.
Does it need a TypeSafe key?
No for the offline keyword fallback; live classification uses TYPESAFE_API_KEY.
Does it execute tasks?
Not yet; it recommends a path and leaves execution to your own handlers.
Related projects
Orchestrator that probes which coding agents are installed and logged in, then has Jev pick the best one for a task.
Agent skillbetaPythonHooks for Claude Code, Codex and pi that use Jev to suggest a skill each turn, plus a rule gate for subagent model routing.
HookbetaJavaScriptClaude Code plugin that asks Jev before each turn for reasoning effort, subagent model, strategy and skill, with a decision log.
Agent skillbetaTypeScriptRouter that puts models, subagents, skills, MCP tools and CLIs in one candidate set and asks Jev one Choice about which handles the step.
MCP serverbetaTypeScriptExperimental plugin wiring Jev into coding-agent harnesses to pick skills, tools, models and browser moves, starting with OpenCode.
Agent skillprototypeTypeScriptPer-turn router: one ~350 ms Jev call picks model tier, effort, tools and skill, with a hard deadline and regex fallback.
Agent skillbetaTypeScript