jevbooks

← All projects

jev-browser-skill

zurfyx/jev-browser-skill · Homepage

Jev-driven browser agent skill for Claude Code and Codex.

Automated README summary

Automatically summarized from the README on 21 Oct 2026; not a run by jevbooks. Run instructions and author-reported outcomes still need verification.

What it is

A reference implementation skill that lets Claude Code and Codex drive a browser toward a stated goal, using Jev to pick each click, type, enter, select, wait or done action on ordinary pages.

How it uses Jev

Each step, SNAPSHOT returns a numbered table of visible controls; buildRequest turns the goal, table, text values and recent actions into named choice questions (operation, click_target, type_target). Jev answers all at once with calibrated probabilities; readDecision validates the pick and browser.mjs executes it.

Primitives:choice

Technique worth stealing

Speculative fan-out: ask the operation and every possible target in one forward pass, so two dependent decisions cost one round trip.

Try it

Clone into ~/.claude/skills/jev-browser or ~/.codex/skills/jev-browser, set TYPESAFE_API_KEY, then ask your agent to use Jev.

View on GitHub

Start using Jev: API key and your first request →

Model judgments from README excerpts

Project signals use repository metadata, an existing summary and the first 700 README characters, not the full project. “Not found” means not detected in that material; it does not establish absence. These are the original judgments with recorded models and dates, not source audits or run results.

Models & evaluation dates

Inclusion gate
jev-1.13.0
Project signals
jev-1.13.0
Category
jev-1.13.0
Pattern tags
jev-1.13.0

Evidence

These signals use repository metadata, an existing summary and at most 700 cleaned README characters. ≥ 0.60 means reported in that excerpt; ≤ 0.40 means not found there. The middle band is inconclusive. None is a source-code audit or a test result. The runnable-tool question excludes research experiments; the recommendation question asks for a learning shortlist, not a quality verdict.

  • Jev is core to the projectFound in material0.85
  • Shows a System One patternunclear0.58
  • Describes handling uncertaintynot found in the excerpt0.07
  • Reports measured numbersnot found in the excerpt0.17
  • Is a runnable toolFound in material0.88
  • Top-10 learning shortlist signalFound in material0.66
  • Is a replica of the modelunclear0.58
  • Solves a common problem (0–2)score on a 0–2 scale0.87 / 2
  • About JevFound in material0.98
Signals
Jev is core to the project0.85Shows a System One pattern0.58Describes handling uncertainty0.07Reports measured numbers0.17Is a runnable tool0.88Top-10 learning shortlist signal0.66Is a replica of the model0.5800.51Solves a common problem (0–2)0.87 / 2012

no-judgment band (0.40–0.60)

Design patterns
Menu0.98Questionnaire0.98Prefetch0.94Inspection0.94Flight recorder0.81Ghostwriter0.46Funnel0.25Viewfinder0.2100.51

no-judgment band (0.40–0.60)

The Jev model and date for each judgment stage are listed above. Card written by DeepSeek V4.1 Flash from the README on 21 Oct 2026.