跳到正文
TakoAPI
返回技能列表

context-clean-up

由 phenomenoner 提供

Coding Agents & IDEs
1.7k下载量由 clawskills.sh 提供
5星标
0浏览量
0 个赞登录后即可为这个技能点赞
—暂无评分
你的评分

简介

Use when: you suspect OpenClaw prompt context is bloating (slow replies, high cost, repeated transcript noise)

描述

该技能的功能

Audits prompt context to identify what is consuming the most space, then produces a ranked list of offenders and a reversible fix plan. No changes are applied automatically. Designed for situations where slow replies, rising costs, or noisy transcripts suggest context has grown too large.

Gives a prioritized, rollback-ready fix plan based on measured context metrics rather than guesswork.

何时使用

  • Diagnosing slow AI replies caused by context bloat
  • Finding oversized tool output clogging session transcripts
  • Identifying cron automation noise accumulating across sessions
  • Spotting bloated bootstrap files like AGENTS.md or SKILL.md
  • Getting a safe trim plan before modifying any config files

示例工作流

输入

User asks: /context-clean-up

智能体
  1. 1

    Determine workspace directory and OpenClaw state directory

  2. 2

    Run the audit script to collect context size metrics across sessions

  3. 3

    Rank offenders by category: tool output dumps, automation noise, bootstrap bloat, ambient skill surface

  4. 4

    Generate a phased fix plan with expected impact and rollback steps for each change

输出

Audit report listing top context offenders, lowest-risk fixes ordered by phase, and verification steps to confirm improvement after changes

安装

clawhub install phenomenoner/context-clean-up

运行 CLI 命令,或把这段提示词粘贴到 Claude Code、Cursor、Codex 或你在用的任意编程智能体里。

链接