Token导航 LogoToken导航

Agent Skill · 研究检索

external-system-safety(外部系统安全)

external-system-safety 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

external-system-safety 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为调用外部服务;许可证为 MIT,暂无第三方扫描结果,按安装量在研究检索类的 73,675 个 Skill 里排第 62,606。

总安装 329周安装 14GitHub Stars 公开资料未说明下载 115核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
MIT

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:external-system-safety(外部系统安全)
来源仓库:https://github.com/duongdev/ccpm
仓库路径:skills/external-system-safety
安装命令:
npx skills add duongdev/ccpm --skill "external-system-safety"
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 duongdev/ccpm --skill "external-system-safety"

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

用于查找、检索和筛选相关信息,快速定位候选结果。适合在 Codex、Claude、Cursor、Gemini CLI 中根据关键词或任务场景使用。通过 npx 命令安装,需指定仓库和技能名称。安装前应确认权限范围和维护状态,避免不必要的网络或文件操作。

external-system-safety 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

来源仓库内路径:skills/external-system-safety

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add duongdev/ccpm --skill "external-system-safety";命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要调用外部服务,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

external-system-safety 怎么安装?

在终端执行 npx skills add duongdev/ccpm --skill "external-system-safety"。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

external-system-safety 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

external-system-safety 需要什么权限?

权限级别是调用外部服务。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。

external-system-safety 是免费的吗?

来源标注的许可证是 MIT。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

external-system-safety 用的人多吗?

公开数据是累计安装 329 次,近一周安装 14 次,下载 115 次。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

external-system-safety 安全吗?

没有查到第三方扫描结果,安全性无法从公开资料判断,安装前请自行审查源码、依赖和命令行为。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 许可证明确为 MIT,商用和分发边界好判断。

先想清楚的地方

  • 没有第三方扫描结果,安全性只能靠自己读源码判断。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 要求有安全扫描结论才肯安装的团队。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
external-system-safety

外部系统安全 · 本页

329—调用外部服务MIT未收录

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

name
external-system-safety
description
Enforces confirmation workflow for EXTERNAL system writes (Jira, Confluence, BitBucket, Slack) ONLY. NEVER activates for Linear operations (Linear is internal tracking). Auto-activates when detecting potential writes to external PM systems (status updates, page creation, PR posts, notifications). Blocks execution and displays exact content that will be written. Requires explicit "yes" confirmation (rejects "ok", "sure", ambiguous responses). All Linear operations execute automatically without confirmation. Works alongside ccpm-code-review to ensure quality before external broadcasts. Provides audit trail of all confirmed operations. Allows batch operations with granular per-item confirmation when needed.
allowed-tools
read-file, grep, browser

External System Safety Guardrails

This skill provides automatic safety enforcement for all operations involving external project management systems.

⚠️ CRITICAL: Linear Exclusion

DO NOT activate this skill for Linear operations. Linear is CCPM's internal tracking system.

NEVER ask for confirmation when:

  • Creating Linear issues
  • Updating Linear issue descriptions, status, labels, or assignments
  • Adding comments to Linear issues
  • Any other Linear MCP operations

This skill ONLY applies to EXTERNAL systems: Jira, Confluence, BitBucket, Slack, etc.


Instructions

⛔ ABSOLUTE RULES - NEVER VIOLATED

Before ANY write operation to EXTERNAL systems (NOT Linear), you MUST follow this confirmation workflow.

1. Detect External System Write Operations

Jira:

  • Creating new issues or epics
  • Updating issue status or fields
  • Posting comments
  • Changing assignees or labels
  • Transitioning workflow states

Confluence:

  • Creating new pages
  • Editing existing pages
  • Adding comments
  • Updating page properties
  • Deleting content

BitBucket:

  • Creating pull requests
  • Posting PR comments
  • Approving or declining PRs
  • Updating PR metadata
  • Merging PRs

Slack:

  • Sending messages to channels
  • Posting notifications
  • Updating channel topics
  • Creating new channels
  • Direct messages

2. STOP and Display Intent

When you detect a potential write operation:

⚠️ EXTERNAL SYSTEM WRITE DETECTED

System: [Jira/Confluence/BitBucket/Slack]
Operation: [Create/Update/Delete/Post]
Target: [Specific item/location]

Details:
---
[Exact content that will be written/posted]
---

This will modify external systems. Proceed? (yes/no)

3. Wait for Explicit Confirmation

Acceptable confirmations:

  • "yes"
  • "confirm"
  • "proceed"
  • "go ahead"
  • "do it"

NOT acceptable (require re-confirmation):

  • Silence
  • "ok" (too ambiguous)
  • "sure" (too casual)
  • Any other response

4. Only Then Execute

After receiving explicit "yes" or equivalent:

  1. Execute the operation
  2. Log what was done
  3. Provide confirmation of success
  4. Show link to updated item if available

Examples

Example 1: Jira Status Update
User: "Update Jira ticket ABC-123 to Done"

Claude: ⚠️ EXTERNAL SYSTEM WRITE DETECTED

System: Jira
Operation: Update Status
Target: ABC-123

Details:
---
Status: In Progress → Done
---

This will modify external systems. Proceed? (yes/no)

User: yes

Claude: ✅ Updated Jira ABC-123 status to Done
Link: https://jira.company.com/browse/ABC-123
Example 2: Confluence Page Creation
User: "Create a Confluence page documenting the new API"

Claude: ⚠️ EXTERNAL SYSTEM WRITE DETECTED

System: Confluence
Operation: Create Page
Target: Engineering Space

Details:
---
Title: "New Payment API Documentation"
Parent: API Documentation
Content: [First 200 characters of content...]
---

This will modify external systems. Proceed? (yes/no)

User: no, not yet

Claude: Understood. I won't create the Confluence page. Let me know when you're ready.
Example 3: Slack Notification
User: "Let the team know the feature is deployed"

Claude: ⚠️ EXTERNAL SYSTEM WRITE DETECTED

System: Slack
Operation: Post Message
Target: #engineering channel

Details:
---
Message: "✅ Payment gateway integration deployed to production. All tests passing."
---

This will modify external systems. Proceed? (yes/no)

User: yes

Claude: ✅ Posted message to #engineering
Example 4: Multiple Operations
User: "Update Jira to Done and notify the team on Slack"

Claude: ⚠️ EXTERNAL SYSTEM WRITE DETECTED

I need to perform 2 write operations:

1. Jira - Update Status
   Target: ABC-123
   Change: In Progress → Done

2. Slack - Post Message
   Target: #engineering
   Message: "Task ABC-123 completed"

Proceed with both? (yes/no)

User: only the jira update

Claude: Got it. I'll update Jira ABC-123 but skip the Slack notification.

⚠️ Confirming: Update Jira ABC-123 to Done? (yes/no)

User: yes

Claude: ✅ Updated Jira ABC-123 to Done
❌ Skipped Slack notification as requested

Always Allowed (No Confirmation Needed)

The following operations NEVER require confirmation:

✅ Read operations:

  • Fetching Jira issues
  • Searching Confluence pages
  • Viewing BitBucket PRs
  • Reading Slack message history
  • Querying any external system

✅ Linear operations:

  • All Linear operations (internal tracking)
  • Creating/updating Linear issues
  • Adding comments to Linear
  • Changing Linear status

✅ Local file operations:

  • Reading/writing files in .claude/ directory
  • Editing codebase files
  • Creating local documentation
  • Git operations (commit, branch, etc.)

✅ Code operations:

  • Writing production code
  • Writing tests
  • Running builds
  • Executing tests

Special Cases

Case 1: User Explicitly Says "Without Confirmation"

If user explicitly requests bypassing confirmation:

User: "Update Jira ABC-123 to Done without asking me"

Still require ONE confirmation for safety:

Claude: I understand you want to skip confirmations, but for safety I need ONE confirmation:

Update Jira ABC-123 to Done? (yes/no)

[After yes, you can skip future confirmations in this session if user requests]
Case 2: Batch Operations

For multiple similar operations:

User: "Update all 5 tickets to Done"

Claude: ⚠️ EXTERNAL SYSTEM WRITE DETECTED

Batch operation: Update 5 Jira tickets to Done
- ABC-123
- ABC-124
- ABC-125
- ABC-126
- ABC-127

Proceed with all? (yes/no/show details)

User: yes

Claude: ✅ Updated all 5 tickets to Done
Case 3: User Seems Confused

If user response is unclear:

User: "maybe"
User: "I guess"
User: "if you think so"

Claude: I need explicit confirmation. Please respond with "yes" to proceed or "no" to cancel.

Integration with CCPM Commands

This skill works alongside CCPM commands that write to external systems:

Commands that trigger this skill:

  • /ccpm:done (creates BitBucket PR, sends Slack notification)
  • /ccpm:sync (updates Jira status)
  • Any command with external system integration

How it works:

  1. User invokes command
  2. Command prepares to write to external system
  3. This skill auto-activates
  4. Confirmation workflow runs
  5. Only after "yes" does command proceed

Error Handling

If external system write fails after confirmation:

Claude: ⚠️ Write operation failed

System: Jira
Error: "Insufficient permissions to update ABC-123"

Possible solutions:
1. Check your Jira permissions
2. Ensure you're on the correct project
3. Verify the ticket exists

Would you like me to retry? (yes/no)

Audit Trail

After any confirmed external write:

Claude: ✅ Operation completed

Audit trail:
- Timestamp: 2025-11-19 14:32:15 UTC
- System: Jira
- Operation: Update Status
- Target: ABC-123
- User confirmed: Yes
- Result: Success

Logged to: .claude/audit-log.json

Why This Skill Exists

Problem: Easy to accidentally write to external systems when:

  • Using agents that bypass command safety checks
  • Misunderstanding what a command does
  • Quickly testing a workflow
  • Copy-pasting commands without reading

Solution: This skill provides an automatic safety layer:

  • ✅ Works even when using agents directly
  • ✅ Catches writes regardless of how they're initiated
  • ✅ Provides clear "what will happen" preview
  • ✅ Gives user final control before execution

Complements existing safety:

  • Commands have built-in safety rules (SAFETY_RULES.md)
  • This skill adds automatic detection layer
  • Works together for defense-in-depth

Reference

For complete safety rules, see: ~/.claude/plugins/ccpm/commands/SAFETY_RULES.md

平台分布

Claude Code

31.34%
按下载量换算36

OpenCode

20.95%
按下载量换算24

Cursor

18.27%
按下载量换算21

Gemini CLI

11.7%
按下载量换算13

kilo

8.68%
按下载量换算10

windsurf

3.65%
按下载量换算4

安全审计

暂无安全审计结果可展示。

权限和风险

调用外部服务

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。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills