本文へスキップ
TakoAPI
スキル一覧に戻る

context-slimmer

sundevilatb 作

Web & Frontend Development
997ダウンロード数はclawskills.shが公開している値です
0閲覧数
0件のいいねサインインするとこのスキルにいいねできます
—まだ評価がありません
あなたの評価

概要

Audit and slim down always-loaded context files (AGENTS.md, TOOLS.md, USER.md, MEMORY.md, HEARTBEAT.md, SOUL.md.

説明

このスキルでできること

Audits always-loaded workspace context files and reduces their token footprint. Measures current token cost across files like AGENTS.md, MEMORY.md, and SOUL.md, then identifies what to move to on-demand skills, remove entirely, or compress. Outputs per-file recommendations with projected savings.

Instead of manually eyeballing files, it measures actual token cost and groups recommendations by action type, so you know exactly what to cut and by how much.

使いどころ

  • Cutting token usage before hitting context window limits
  • Finding duplicate content spread across MEMORY.md and AGENTS.md
  • Moving task-specific config out of always-loaded files into on-demand skills
  • Auditing a workspace after months of accumulated context drift
  • Setting and enforcing lean file size targets for each context file

ワークフロー例

入力

User asks: audit my context files and reduce token usage

エージェント
  1. 1

    Run bash scripts/measure.sh --audit to measure token cost per file

  2. 2

    Compare each file's current size against lean targets from the reference table

  3. 3

    Identify content to move to on-demand skills, such as task-specific protocols or betting config

  4. 4

    Flag outdated, duplicated, or overly verbose sections for removal or compression

  5. 5

    Output a report with current size, projected size, and savings per file

出力

Per-file audit report showing token counts, specific move/remove/compress recommendations, and total projected savings

インストール

clawhub install sundevilatb/context-slimmer

CLIコマンドを実行するか、Claude Code、Cursor、Codexなどお使いのコーディングエージェントにプロンプトを貼り付けてください。

リンク