#268268th of 572 in Skills, plugins and prompts

chisle

Ruleset that cuts coding agents' output tokens and trims tool output

Releases and problems feed (RSS)

Stars
1.2k
License
MIT
Last commit
Nov 2026
Last release
Nov 2026
Language
JavaScript
Stars, last 20 weeks+1,194
1 Jul 2026: 2 stars8 Jul 2026: 2 stars15 Jul 2026: 89 stars22 Jul 2026: 93 stars29 Jul 2026: 98 stars5 Aug 2026: 98 stars12 Aug 2026: 99 stars19 Aug 2026: 100 stars26 Aug 2026: 101 stars2 Sept 2026: 101 stars9 Sept 2026: 110 stars16 Sept 2026: 426 stars23 Sept 2026: 510 stars30 Sept 2026: 577 stars7 Oct 2026: 630 stars14 Oct 2026: 662 stars21 Oct 2026: 692 stars28 Oct 2026: 722 stars4 Nov 2026: 739 stars11 Nov 2026: 758 stars18 Nov 2026: 1,196 stars

2 on 1 Jul 20261,196 on 18 Nov 2026

From GitHub's daily star counts.

Checked on GitHub.

Overview

Chisle installs a terse-output ruleset and a tool-output compressor into coding agents such as Claude Code, Gemini CLI, Codex, OpenCode, Copilot CLI and Cursor. It strips ANSI codes, collapses repeated lines and elides long command output, and it can wrap MCP servers to trim their output. The README reports 29% fewer billed output tokens and 156/156 accuracy versus 154/156 with no ruleset.

Who it is for: Developers paying per token for Claude Code or similar coding agents

Strengths

  • One `npx chisle` command detects and configures several agents; --dry-run and --uninstall exist
  • Command-aware filters for npm test, go test and git keep errors and failures
  • Raw benchmark transcripts are committed, including cases where it loses
  • Adds /chisle-review and /chisle-audit commands plus MCP output wrapping

Weaknesses

  • Benchmarks are self-published: 25 tasks on Claude Code, 156 accuracy checks
  • Savings on git output are small: git status -30%, git log -15%
  • Compression elides the middle of long output; the full text is saved to a file
  • Targets output tokens, not the cost of reading a large repo

What it needs

  • no GPU
  • Needs Node.js (npx)
  • Models: Claude Code, Gemini CLI, Codex, OpenCode, Copilot CLI

What changed

  1. “BREAKING: one mode. lite, full and ultra are gone. Three dials on a”

  2. “BREAKING: one mode. lite, full and ultra are gone. Three dials on a”

  3. “BREAKING: one mode. lite, full and ultra are gone. Three dials on a”

From the release notes. No flag means the note does not say so, not that the upgrade is safe.

Also in Skills, plugins and prompts

See all 572
Also in Skills, plugins and prompts
RankProjectScore
1claude-memPersistent cross-session memory plugin for Claude Code and other coding agents118.3k stars+16,740 (+16%) in 30 daysApache-2.097 out of 100
2impeccableDesign skill with 24 slash commands for AI coding agents' frontend output Live demo ↗ (opens in a new tab)113.1k stars+29,544 (+35%) in 30 daysApache-2.097 out of 100
3gstackMarkdown slash-command skills that give Claude Code planning, review, QA and ship roles142.1k stars+5,409 (+4.0%) in 30 daysMIT96 out of 100
#4rufloAgent harness adding swarms, memory and hooks to Claude Code and Codex Live demo ↗ (opens in a new tab)82.1k stars+4,953 (+6.4%) in 30 daysMIT96 out of 100
#5eccAgents, skills, hooks and memory bundle for Claude Code and other harnesses306.8k stars+22,797 (+8.0%) in 30 daysMIT95 out of 100