Token导航 LogoToken导航TokenDH.com
研究检索external-servicegithub未标认证来源可访问clear审计未展示

writing-agents写作 Agent

Agent Skill

用于辅助文档、README、Markdown、说明文和内容稿件的整理与改写。它适合让 Agent 提炼结构、补齐章节、统一术语、检查链接或把零散材料整理成可读文档。使用时应保留项目已有事实、命令和路径,不要把未确认的信息写成确定结论;涉及对外文案时,还需要控制语气,避免过度营销或夸大能力。

总安装

250

周安装

10

GitHub Stars

公开资料未说明

下载量

81
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

MIT

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:writing-agents(写作 Agent)
来源仓库:https://github.com/ian-pascoe/dotfiles
仓库路径:skills/writing-agents
安装命令:
npx skills add ian-pascoe/dotfiles --skill "writing-agents"
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

AgentSkills.tonpx skills
npx skills add ian-pascoe/dotfiles --skill "writing-agents"

简介

辅助文档、README 和内容稿件的整理与改写,提升可读性。

  • 可提炼结构、统一术语、检查链接并整合零散材料。
  • 保留已有事实和路径,避免将未确认信息写成确定结论。
  • 对外文案需注意语气控制,防止过度营销或夸大能力。
  • writing-agents 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

SKILL.md

name
writing-agents
description
Use when creating, editing, or reviewing agent configuration files (AGENT.md or opencode.json agents). Covers frontmatter options, mode selection, tool permissions, and common patterns.

Writing Agents

Overview

Agents are specialized AI assistants with custom prompts, models, and tool access. Configure via markdown files or JSON.

When to Use

  • Creating a new agent for specific tasks
  • Editing agent prompts, permissions, or tool access
  • Choosing between primary vs subagent mode
  • Setting up task permissions for orchestration
  • Reviewing agent configuration quality

Don't use for: skill files (use writing-skills), system prompts without agent context, or MCP server config.

File Locations

~/.config/opencode/agent/<name>.md     # Global agents
.opencode/agent/<name>.md              # Project-local agents
opencode.json → agent.<name>           # JSON config (either location)

Filename becomes agent name (e.g., review.mdreview agent).

Frontmatter Reference

FieldRequiredValuesNotes
descriptionYesstringBrief description for agent selection
modeNoprimary, subagent, allDefault: all
modelNoprovider/model-idInherits from parent if unset
temperatureNo0.0-1.00.0-0.2 focused, 0.6-1.0 creative
maxStepsNointegerLimit agentic iterations
hiddenNobooleanHide from @ autocomplete (subagents only)
disableNobooleanDisable the agent
toolsNoobjectEnable/disable specific tools
permissionNoobjectConfigure ask/allow/deny per tool

Mode Selection

ModeUse Case
primaryMain agents user switches between (Tab key)
subagentInvoked by primary agents or via @ mention
allAvailable in both contexts (default)

Tool Configuration

tools:
  write: false # Disable write tool
  edit: false # Disable edit tool
  bash: false # Disable bash tool
  mymcp_*: false # Wildcards for MCP tools

Permission Configuration

permission:
  edit: deny # deny | ask | allow
  bash:
    "*": ask # Default for all commands
    "git status": allow # Specific command override
    "git push*": ask # Glob patterns supported
  webfetch: deny
  task:
    "*": deny # Block all subagent invocation
    "my-helper-*": allow # Allow specific subagents

Rule evaluation: Last matching rule wins. Put * first, specifics after.

Markdown Template

---
description: <what it does and when to use>
mode: subagent
model: anthropic/claude-sonnet-4-20250514
temperature: 0.1
tools:
  write: false
  edit: false
  bash: false
permission:
  edit: deny
---

<system prompt content>

Focus on:

- Point 1
- Point 2

JSON Template

{
  "agent": {
    "my-agent": {
      "description": "What it does",
      "mode": "subagent",
      "model": "anthropic/claude-sonnet-4-20250514",
      "prompt": "{file:./prompts/my-agent.txt}",
      "tools": {
        "write": false,
        "edit": false
      },
      "permission": {
        "bash": {
          "*": "ask",
          "git diff": "allow"
        }
      }
    }
  }
}

Common Patterns

Read-Only Reviewer

tools:
  write: false
  edit: false
  bash: false

Restricted Bash Access

permission:
  bash:
    "*": ask
    "git status": allow
    "git log*": allow
    "git diff*": allow

Orchestrator with Task Limits

permission:
  task:
    "*": deny
    "my-workers-*": allow

Common Mistakes

MistakeFix
Missing descriptionAlways provide clear description
Using hidden on primary agenthidden only works on subagents
Expecting * to be evaluated lastLast rule wins - put * first
Setting model on subagentSubagents inherit parent model if unset
Narrative prompt styleFocus on role and constraints, not conversation

Validation Checklist

  • [ ] Description clearly states purpose and trigger conditions
  • [ ] Mode matches intended usage (primary vs subagent)
  • [ ] Tools restricted appropriately for the role
  • [ ] Permissions use correct ask/allow/deny values
  • [ ] Bash permission rules ordered correctly (* first)
  • [ ] Prompt focuses on role, constraints, and focus areas

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

04

需要参考平台分布和安装热度时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

补充不同宿主或平台的使用分布数据

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

OpenCode

26.77%
按下载量换算22

Claude Code

23.04%
按下载量换算19

windsurf

20.26%
按下载量换算16

Codex

12.16%
按下载量换算10

Antigravity

8.44%
按下载量换算7

Gemini CLI

3.3%
按下载量换算3

安全审计

暂无安全审计结果可展示。

权限和风险

external-service

该 Skill 可能调用第三方服务、云服务或外部模型 API,使用前需要确认账号、额度、数据发送范围和服务条款。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills