一个只读的 Codex 技能,通过分析本地遥测数据提供诊断和活动回顾,帮助用户理解和优化 Codex 使用情况。

Stars

249

7 天增长

暂无数据

Fork 数

7

开放 Issue

0

开源协议

MIT

最近更新

2026-07-19

AI 仓库情报摘要
FR-AI / ANALYSIS

为什么值得关注

它提供了一个注重隐私、开源的 Codex 审计工具,直接在用户机器上运行,生成结构化报告,将证据与解释分离,并推荐可逆的优化步骤。

适合谁使用

  • 希望减少 token 浪费的 Codex 高级用户
  • 管理多个插件和 MCP 配置的开发者
  • 注重隐私、希望在不外传数据的情况下分析使用情况的用户

典型使用场景

  • 回顾过去 1 到 90 天的 Codex 活动
  • 识别哪些插件消耗了最多的工具运行时间或序列化输出
  • 生成 Markdown 或 JSON 格式报告用于团队审计或个人优化
  • 制定安全且可逆的清理计划,如禁用不用的插件或调整配置

项目优势

  • 以只读模式运行(使用 sqlite3 -readonly),确保不修改任何数据
  • 明确区分观察到的证据、解释和未知信息,提高报告清晰度
  • 支持快速测量和详细的周期回顾,时间范围灵活
  • 全面保护隐私,提示词、回复、路径和密钥均保留在本地

使用前须知

  • 仅兼容 macOS 和类 Unix 环境,不支持 Windows
  • 依赖版本特定的遥测模式,Codex 更新可能导致兼容性问题
  • 无法提供计费归属或实时监控功能

README 快速开始

Codex Hygiene

A compact, read-only Codex skill for current-context diagnostics and private local activity reviews.

The original measurement returns small counts from recent telemetry. An explicit 1-90 day review returns Markdown or JSON that describes how Codex was used, which sources support the report, and where the evidence ends.

What It Does

  • Runs the original compact measurement for recent tool-list and per-thread token telemetry.
  • Builds a period review with source coverage and supported prior-window comparisons.
  • Separates observed evidence, interpretation, and unknowns.
  • Enriches period reviews with task timing, relative output weight, verification commands, explicit compactions, and observed SKILL.md reads.
  • Reviews the current Codex profile, including every installed plugin, MCP state, Apps availability, and project-stanza counts.
  • Suggests scoped, reversible hygiene steps after measurement.

Install

Clone this repo into the current user-level Codex skills folder:

mkdir -p "$HOME/.agents/skills"
git clone https://github.com/sunflower-of-parchman/codex-hygiene.git \
  "$HOME/.agents/skills/codex-hygiene"

Then invoke $codex-hygiene. Codex usually detects newly installed skills automatically. Restart Codex if the skill has not appeared.

Quick Measurement

From a shell:

SKILL_DIR="$HOME/.agents/skills/codex-hygiene"
"$SKILL_DIR/scripts/measure_codex_context.sh"

Use a specific window or thread id:

"$SKILL_DIR/scripts/measure_codex_context.sh" 5
"$SKILL_DIR/scripts/measure_codex_context.sh" 30
"$SKILL_DIR/scripts/measure_codex_context.sh" 5 

The script prints compact counts. Full logs, configs, tool schemas, secrets, and environment values stay private.

Activity Review

Choose an explicit lookback from 1 to 90 days:

SKILL_DIR="$HOME/.agents/skills/codex-hygiene"
python3 "$SKILL_DIR/scripts/codex_activity_review.py" --days 1

Use any other window or generate machine-readable output:

python3 "$SKILL_DIR/scripts/codex_activity_review.py" --days 14 --format json

The review reads compact SQLite telemetry and enriches it from rollout records inside CODEX_HOME. Enrichment runs by default at every candidate size and reads files backward only to the prior comparison boundary. It adds task timing, relative serialized tool-output weight, verification-command counts

项目描述

Codex skill for auditing and tuning Codex Desktop context/tool surfaces

相关仓库与替代方案

根据分类、Topic 和编程语言匹配的相似项目。

powerycy
精选
powerycy GitHub avatar

goutoujunshi

An AI relationship coach that provides emotional support, relationship science, and actionable strategies through a multidisciplinary knowledge base.

Python
1,174
Kappaemme-git
Kappaemme-git GitHub avatar

codex-first-customer-finder-skill

A Codex skill that converts a startup URL or product idea into a qualified shortlist of potential first customers by analyzing public pain, demand, and timing signals.

Python
932
0xsline
0xsline GitHub avatar

OpenChatCut

OpenChatCut is an open-source, local-first AI video editor with a real multitrack timeline, conversational agents, and MCP support, enabling AI-assisted editing that remains fully editable.

TypeScript
579