colly-rd-risk-mcp-installer

SKILL运维和基础设施
运维和基础设施 未标注权限 许可证不明确
安装这个技能

安装xsm本地环境的mcp 研发风险平台

一句话选型结论

按资料字段整理

colly-rd-risk-mcp-installer是运维和基础设施类的 Agent Skill,安装方式未标注,权限级别未标注,许可证不明确,有可直接复制的安装命令。技能本质是给模型读的说明书,效果取决于模型与客户端;装完先用一个小任务验证,再放进正式流程。

分类运维和基础设施 安装方式未标注 权限级别未标注权限 许可证许可证不明确
累计安装 未标注 本周 未标注
GitHub 星标 未标注 仓库 未标注
下载量 未标注 技能市场下载数
权限级别 未标注权限 安装前请确认风险范围
许可证 许可证不明确 未说明

安装

未标注
在支持 Skills 的客户端里执行,装完问一句「你有哪些技能」验证
skillhub install user_a31ac21c/risk-test
  1. 1
    确认客户端支持

    确认你的客户端支持 Skills(Claude、Codex、Cursor 等),并升级到较新的版本。

  2. 2
    拿到技能文件

    执行上面的命令安装;也可以把仓库里的 SKILL.md 手动放进客户端的 skills 目录。

  3. 3
    刷新技能列表

    重新打开客户端,或刷新技能列表,让新放进来的技能被加载。

  4. 4
    验证是否生效

    问一句「你有哪些技能」,确认列表里出现它,再拿一个小任务试跑。

技能档案

运维和基础设施
字段来自技能仓库与公开索引,中文说明由本站标注
作者 / 维护方user_a31ac21c仓库 未标注 · 版本 未标注
宿主客户端未标注技能可被这些客户端加载
安装方式未标注安装包 community
权限级别未标注权限安装前请确认风险范围
许可证未说明许可证不明确,商用前请自行确认条款
仓库可用性available技能仓库当前是否可访问
最近核验未标注上游最后核验时间 · 数据更新 1791625307
正文体量6.2 KBSKILL.md 原文大小,越大说明说明书写得越细

典型使用场景

3 类
按分类整理的常见用法,具体以 SKILL.md 为准
查服务与日志

按现象定位服务状态与日志线索。

写部署与巡检脚本

把重复的部署、巡检写成脚本。

故障排查

按步骤排查网络、依赖与配置。

选择建议

装之前先看这几条
按条目字段给出的判断,不替你决定用不用

值得选的地方

  • 有可直接复制的安装命令,不用自己翻文档找入口。

先想清楚的地方

  • 没有标注权限级别,实际能做什么要装好后确认。
  • 许可证不明确,商用或二次分发前先到仓库确认条款。
  • 没有做站内核验,内容以作者仓库为准。
  • 技能只是说明书,实际效果取决于模型与客户端,装完先用小任务试一遍。

更适合

  • 需要在运维和基础设施场景里给模型补一套做法的人。
  • 愿意先读一遍 SKILL.md、确认它会让模型做什么的人。

不太适合

  • 对许可证要求很严、必须明确可商用的人。
  • 只想装完就自动变强、不打算看说明书的人。

同类对比

6 个同类
同分类里挑最接近的几条,字段以各自详情页为准
技能权限级别安装方式累计安装宿主客户端安装命令
colly-rd-risk-mcp-installer当前 未标注权限 未标注 未标注 未标注 有
Azure 部署 调用外部服务 GitHub 仓库 664.2 万 Codex、Claude、Cursor 有
Azure diagnostics 问题管理 调用外部服务 GitHub 仓库 657.6 万 Codex、Claude、Cursor 有
Azure aigateway 部署 调用外部服务 GitHub 仓库 655.3 万 Codex、Claude、Cursor 有
Azure AI 搜索 调用外部服务 GitHub 仓库 651.4 万 Codex、Claude、Cursor 有
Azure Cloud migrate 部署 调用外部服务 GitHub 仓库 650.7 万 Codex、Claude、Cursor 有

SKILL.md 全文

6.2 KB
模型实际读到的说明书原文,来自技能仓库

Colly R&D Risk MCP Installer

Scope

This Skill supports Windows clients that can reach 192.168.30.136 on the trusted LAN or approved VPN. The fixed MCP endpoint and OAuth resource are both https://192.168.30.136/mcp.

The shared layer performs LAN, pinned-CA, current-user trust-store, and HTTPS health checks. It does not assume a client product or write an unknown client's configuration.

Do not adapt this workflow to DEAP or another vendor-cloud executor: those services cannot directly route to this private endpoint and may reject it through SSRF protection.

Capability routing

Select a capability mode, not a product name:

  1. Use ConnectionProfile when the consuming client is unknown, undocumented, or only needs the server/OAuth/TLS contract. This is the safe fallback.
  2. Use NativeHttpOAuth only when the client supports Streamable HTTP, OAuth discovery, DCR, PKCE S256, RFC 8707 resource, refresh tokens, and an explicit-port loopback HTTP callback.
  3. Use StdioOAuthProxy when the client supports local STDIO configuration but cannot complete remote interactive OAuth. This generates a pinned third-party local proxy configuration and requires user review before import.
  4. Use Preflight for network, CA, and HTTPS diagnosis without producing a client connection artifact.
  5. Use Codex, WorkBuddy, or TraeWork only as optional convenience presets when the actual host is known and the user wants its specific configuration mutation. Never choose a preset merely because its executable or directory exists.

Read Local environment and troubleshooting before running a mode. Then read only the selected mode guide: Connection profile, Native HTTP OAuth, or STDIO OAuth proxy. Product-specific behavior remains in Codex, WorkBuddy, and TraeWork.

Execution workflow

Use the public dispatcher from a foreground PowerShell session:

powershell.exe -ExecutionPolicy Bypass -File "<skill-directory>\scripts\configure_colly_mcp.ps1" -Mode ConnectionProfile
powershell.exe -ExecutionPolicy Bypass -File "<skill-directory>\scripts\configure_colly_mcp.ps1" -Mode NativeHttpOAuth
powershell.exe -ExecutionPolicy Bypass -File "<skill-directory>\scripts\configure_colly_mcp.ps1" -Mode StdioOAuthProxy

Unknown clients default to ConnectionProfile at the Skill decision layer; the script itself still requires an explicit -Mode so side effects are auditable. Do not claim that the client-neutral profile is a standardized MCP import format. MCP standardizes protocol and OAuth behavior, not every client's configuration path, JSON/TOML schema, CLI, approval UI, or reload lifecycle.

The user's request to install this specific MCP authorizes adding its pinned public CA to the current-user trust store and writing artifacts under the dedicated local Colly adapter directory. It does not authorize guessing or modifying an unknown client's configuration, installing another product, inserting static credentials, or claiming generated artifacts have already been imported or authenticated.

The bootstrap compares the process identity profile with the target interactive Windows profile. If they differ, a real run must stop before certificate or artifact mutation; rerun through the host approval mechanism in a foreground process as the target Windows user.

The user must personally select or scan the DingTalk account and approve the listed scopes. Never automate account selection, enter credentials, or treat a browser page load as successful authorization.

If an entry already exists with a conflicting endpoint, transport, generated artifact, or static authorization header, stop without printing credential values or overwriting it automatically. Never create a Bug, task, Story, notification, or configuration change as an installation test.

Authorization boundary

Request only the ordinary-user baseline scopes:

risk:read,zentao:bug:create,zentao:task:create,zentao:story:create

Do not request risk:write, risk:notify, or risk:admin during normal installation. DingTalk OAuth identifies the MCP user; personal ZenTao authorization is a separate later step used only for a confirmed Bug, task, or Story creation. Never request or handle passwords, cookies, OAuth tokens, API tokens, or browser sessions.

Certificate rules

  • Download only the fixed LAN CA URL over a direct connection. Do not follow redirects, accept more than 16 KiB, or treat plaintext transport as trusted.
  • Trust the public root only after its pinned file SHA-256 and certificate DER SHA-256 match and the certificate is a currently valid CA.
  • Import into Cert:\CurrentUser\Root; never search for or distribute Caddy private keys.
  • Never use insecure TLS, copied Bearer tokens, browser cookies, renamed certificate assets, or embedded PEM/Base64 as a workaround.
  • If validation fails, stop and require the administrator to repair the LAN endpoint or publish updated fingerprints.

Completion response

Report the selected capability mode and separately state network, CA trust, artifact generation or preset mutation, OAuth, and same-session read-only verification.

Use these states precisely:

  • prepared: connection artifacts exist;
  • imported: the consuming client accepted them;
  • authenticated: the user's loopback OAuth completed;
  • verified: that exact client session successfully called read-only get_status.

Then summarize the actual authorized capability groups from Capabilities, noting that ZenTao writes still require the applicable personal authorization and server-side confirmation contract.

常见问题

5 问
答案由本站依据条目字段整理
colly-rd-risk-mcp-installer怎么安装?

在支持 Skills 的客户端里执行:skillhub install user_a31ac21c/risk-test。安装方式为未标注。装完刷新技能列表或重开客户端。

colly-rd-risk-mcp-installer支持哪些客户端?

资料没有标注宿主客户端,但只要客户端支持 Skills(Claude、Codex、Cursor 等),一般都能加载。

colly-rd-risk-mcp-installer权限高吗,安全吗?

权限级别为未标注权限。安装前先看它会读取或写入哪些内容。

colly-rd-risk-mcp-installer收费吗?

本站不代收费用,也没有标注具体价格。是否收费取决于技能本身和所用的模型或客户端,正式使用前按作者说明确认。

colly-rd-risk-mcp-installer用什么许可证?

许可证标注为 未说明(许可证不明确)。商用或二次分发前请到仓库确认条款原文。

热门专题

按场景挑技能
全部专题 →

技能内容以作者仓库为准;分类、权限级别与许可证状态的中文说明为中文整理,最近核验 未标注。