jevbooks

← All projects

jev-pruner

tamaratran/jev-pruner

Claude Code plugin trims long Bash output with Jev before the model sees it.

What it is

A Claude Code plugin for developers that uses TypeSafe's Jev to prune noisy Bash stdout after a command runs but before the result reaches the main LLM, reducing tokens carried into later turns without generating a summary.

How it uses Jev

A tool.call hook wraps the Bash tool's next() result. Jev receives { context, task, history, command, chunks } and one noul question per chunk: does any line in this chunk need to remain available? Chunks scoring at least keepThreshold or above 0.1 are kept; dropped runs become [N lines omitted].

Primitives:noul

Technique worth stealing

Gate output by an estimated-token floor, then ask Jev one noul question per chunk against partitioned history segments.

View on GitHub

judged by Jevjev-1.13.0

Evidence

Each line is one question put to Jev about the README. ≥ 0.60 reads as yes, ≤ 0.40 as no; in between Jev is not making a call.

  • Jev-centricunclear0.40
  • Shows a System One patternno0.29
  • Handles uncertaintyno0.06
  • Measuredno0.05
  • Runnableyes0.98
  • Worth recommendingno0.31
  • Model replicano0.03
  • Problem scopescore on a 0–2 scale1.09
  • About Jevyes0.97

Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.