ego-jev
Agent skill where Ego Lite drives the browser and Jev resolves semantic choices: targets, filtering, ranking and evidence checks.
Overview
This ego-jev skill splits browser work between an agent, Ego Lite and Jev. The agent keeps the goal, the plan and any writing; Ego Lite drives a real Chromium browser (snapshots, DOM actions, input, tabs, uploads, screenshots); Jev resolves semantic choices such as which observed link, button or card matches the intent, how to categorise or rank items, and whether page text is real evidence. Exact work such as prices, arithmetic, counts, sorting and date comparison stays in local code, and visual judgments stay with the agent's vision.
The README stresses that speedups come from fewer planner turns and batched judgments, not from routing every browser action through an API, and that known actions need no inference while semantic choices must go through Jev rather than being relabelled as navigation. It is listed on skills.sh.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which observed link, button or card matches the user's intent? | — |
| Score | How relevant or high-quality is each item? | — |
| Yes / no | Is this page text real evidence for the claim? | yes · no |
How it works
- Agent plans and observes the page via Ego Lite
- Semantic choices are batched to Jev
- Local code handles exact comparisons
- Ego Lite executes verified actions
Strengths
- Clear division of labour
- Batched judgments reduce planner turns
- Exact work kept deterministic
Caveats
- Relies on the agent following the contract
- Requires Ego Lite and a TypeSafe key
Getting started
npx skills add ZHUBoer/ego-jev.At a glance
- Type
- Project
- Status
- beta
- Language
- Markdown + TypeScript
- Runtime
- agent skill for Ego Lite
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Complete browser tasks with Ego Lite and actively call Jev for semantic target selection, filtering, ranking, classification and text evidence judgments.
logicrw/awesome-jev-projects
Frequently asked questions
What does Jev handle?
Semantic selection, classification, per-item relevance and evidence checks.
What stays in code?
Prices, arithmetic, counts, sorting, dates and exact matching.
How do I install it?
npx skills add ZHUBoer/ego-jev.
Related projects
MCP server and libraries that drive Chrome over CDP and hand a decision model like Jev a bounded menu of page actions.
ProjectproductionTypeScript- Jev Browser★ 335
Headless browser automation via MCP server, CLI or library: Jev picks one action per step and scores goal-met and stuck probabilities.
MCP serverbetaTypeScript Claude Desktop MCP extension that drives Brave or Chrome with Jev choosing on-page actions and a guard freezing irreversible clicks.
MCP serverbetaTypeScriptChrome MCP server with 50 tools that drives your everyday browser without losing logins, where Jev or a local decider picks each step.
MCP serverbetaStealth browser agent where Jev decides each step in about 0.3 s and CloakBrowser carries it out; MCP server, CLI and Python API.
MCP serverbetaPythonCodexQA browser skill that indexes in-page controls and has Jev pick the operation and target each step, with no screenshots needed.
Agent skillbeta