Reduce coding-agent token waste with Chisle
Use Chisle to add concise coding rules, YAGNI-first defaults, and supported tool-output compression across multiple coding agents.
npx skills add agentskillexchange/skills --skill reduce-coding-agent-token-waste-with-chisle
Use Chisle when an operator wants a coding agent to spend fewer output and context tokens without changing the underlying model or editor. The workflow is concrete: install Chisle, let it detect supported agents, enable the ruleset for terse prose and YAGNI-first implementation, use supported hooks or toggles for input-side compression where available, preview changes before applying them, and update or uninstall cleanly when the policy is no longer wanted.
What this skill actually does
Invoke this instead of manually telling every agent to be concise when the same repository is used across Claude Code, Pi, Cursor, Windsurf, Cline, Kiro, Codex, Gemini, Copilot, OpenCode, Hermes, or similar hosts. It is strongest when tool output repeatedly floods the context window or when a team wants a reversible, measured agent-efficiency policy across several coding agents.
This is not a generic prompt-pack or token dashboard listing. The scope boundary is coding-agent efficiency: terse response defaults, YAGNI-first implementation guidance, selected tool-output compression, per-agent installation/update/uninstall flows, and published benchmark comparisons. Do not stretch it into general cost accounting, model routing, or unrelated prompt engineering.
Inputs and prerequisites: Node.js 18+; Chisle npm package; one or more supported coding agents; Claude Code or Pi for the richest toggling and input-side compression behavior.
Setup notes: Run npx chisle to start the installer. Preview with npx chisle –dry-run, scope installation with –only when needed, update with npx chisle@latest –update, and remove with npx chisle –uninstall.
Source and verification boundary: use https://github.com/JayPokale/Chisle and https://chisle.jaypokale.me as canonical sources. Check upstream benchmark notes and per-agent support before claiming token savings in a specific environment.
Framework fit: publish as Multi-Framework because upstream explicitly supports many coding agents and the workflow is an agent-agnostic efficiency layer, not a Codex-only or Claude-only integration.