Chisle
A tool that cuts an agent's token bill on three axes: terser prose, less speculative code, compressed tool output
Medium risk
We rate an entry medium when the tool runs code, makes network calls or reads project files. Check what exactly it does before installing.
Why this level
- Installs hooks that run local code on session start and after tool calls
- Changes agent configuration on your machine
Install
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/chisleDetects the agents on your machine, checks the risk and pins the version.
Other ways to install
Install the tool
npm install -g chisleInstall Chisle: npx chisle, and it detects your agents and wires the ruleset. For Claude Code you can use the plugin: claude plugin marketplace add JayPokale/Chisle, then claude plugin install chisle@chisle.
Other ways from the author
npx chisleDetects installed agents and wires the ruleset into each, and npx chisle --uninstall reverts everything.
This is third-party code. Review the repository files before installing.
What it does
Chisle finds installed agents with one command and wires a ruleset into each. The rules make answers terser without filler and push code toward the least that is needed, with no extra abstractions. In Claude Code and Pi it also compresses tool output on the input side: it scrubs, elides and dedups text that would otherwise be billed on every later step of the session. It has a dry run, a savings counter and a clean uninstall. It has no dependencies and publishes benchmarks, including the runs it loses.
Who it is for. For developers who want to shrink their coding agent's token spend.
Good fit when
- The agent gives verbose answers and extra abstractions
- Tool output bloats the context and is billed over and over
- You want to set up several agents with one command
Not a fit when
- You want thorough explanations and detailed options rather than the tersest answer
- The agent is not on the supported list
Example request
Set up Chisle for my agents and show what changes before applyingLimitations
Tool-output compression works in Claude Code and Pi, while other agents get only the ruleset. A plain run does not upgrade what is already installed, so an update needs a separate flag. Some agents keep the rule file inside the repository, so you repeat the setup in each such project.
How to disable. Run npx chisle --uninstall, which puts everything back for every configured agent.
Security check
- Installs hooks that run local code on session start and after tool calls
- Changes agent configuration on your machine
README in short
The README describes a tool that lowers token spend on three fronts at once: terser prose, least-needed code, and input-side compression of tool output. It installs with a single npx chisle command that detects agents and wires the ruleset, with a Claude Code plugin path as well. It ships benchmarks on live tasks and shows its worst cases honestly, where the tool wrote more than the bare model. It claims support for eleven agents and zero dependencies. It has a preview, a savings counter and an uninstall.
FAQ
How do I preview the changes?
npx chisle --dry-run prints every file the tool would touch and changes nothing.
How much did it save?
npx chisle --stats shows the cumulative savings from tool-output compression.
Related
A skills library that gives coding agents a development process: brainstorming, planning, TDD, subagents and code review
Skills for real engineers by Matt Pocock
Skills For Real Engineers
Small composable skills for engineering with agents: plan grilling, TDD, bug diagnosis, code review and architecture
GitHub toolkit for spec-driven development: the specify CLI adds agent commands and skills to a project, from principles to implementation
Reference MCP servers
Model Context Protocol servers
Official reference MCP servers: Filesystem, Fetch, Git, Memory, Sequential Thinking, Time and Everything