Agent Skillsfanko1217/codex-work-starter › codex-work-starter

codex-work-starter

GitHub

帮助非程序员将日常办公任务转化为安全、可检查的Codex工作流,涵盖邮件、文档处理及会议准备,强调手动验证与隐私保护。

codex-work-starter/SKILL.md fanko1217/codex-work-starter

Trigger Scenarios

非程序员希望使用Codex处理日常工作 需要建立跨应用的主管助手工作流

Install

npx skills add fanko1217/codex-work-starter --skill codex-work-starter -g -y
More Options

Non-standard path

npx skills add https://github.com/fanko1217/codex-work-starter/tree/main/codex-work-starter -g -y

Use without installing

npx skills use fanko1217/codex-work-starter@codex-work-starter

指定 Agent (Claude Code)

npx skills add fanko1217/codex-work-starter --skill codex-work-starter -a claude-code -g -y

安装 repo 全部 skill

npx skills add fanko1217/codex-work-starter --all -g -y

预览 repo 内 skill

npx skills add fanko1217/codex-work-starter --list

SKILL.md

Frontmatter
{
    "name": "codex-work-starter",
    "description": "Convert everyday email, calendar, meeting, spreadsheet, document, study, and website work into safe, checkable Codex workflows. Use when a non-programmer wants to start using Codex for real work, set custom instructions, organize projects and long threads, build a chief-of-staff workflow, prepare meetings or papers, clean spreadsheets, create dashboards, or schedule recurring briefs."
}

Codex Work Starter

Start with one real task, prove it works manually, then make it reusable.

Read references/prompts.md for the complete prompt library.

Setup

  1. Write short custom instructions describing the user's role, preferred language, candor, formatting, and privacy boundaries.
  2. Connect only the apps required for the task.
  3. Use one project for one major responsibility and one long thread for one continuing task.
  4. Create a dedicated chief-of-staff thread for cross-app work.

Safety boundary

  • Reading, organizing, and drafting can proceed within the authorized scope.
  • Sending, unsubscribing, inviting, deleting, publishing, purchasing, or changing permissions must stop for human confirmation unless the user explicitly authorizes that exact action.
  • Keep useful source links while removing unnecessary sensitive detail.
  • Never expose passwords, identity numbers, payment data, or secret keys in a report or recording.

Workflow

  1. Run the prompt manually.
  2. Inspect source coverage and the result.
  3. Correct tone, omissions, and privacy leaks.
  4. Save the successful method as a prompt or skill.
  5. Only then turn it into a recurring task.

Verification

For each task, report what was read, what changed, what was not changed, how the result was checked, and which important action still awaits confirmation.

Version History

  • 0bb7bad Current 2026-08-05 14:50

Metadata

Files
0
Version
0bb7bad
Hash
47c32eaa
Indexed
2026-08-05 14:50

Home - Wiki
Copyright © 2011-2026 iteam. Current version is 2.155.2. UTC+08:00, 2026-08-06 14:17
浙ICP备14020137号-1 $Map of visitor$