Token导航 LogoToken导航

Agent Skill · 开发

rfc(射频卡)

rfc 用于处理 GitHub 仓库、Issue、Pull Request 和代码协作信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要围绕仓库状态、代码变更或协作事项进行整理时使用。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。

rfc 是开发类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为需要联网;许可证信息还不明确,第三方扫描已通过,按安装量在开发类的 21,746 个 Skill 里排第 20,160。

总安装 238周安装 10GitHub Stars 公开资料未说明下载 83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
unknown

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:rfc(射频卡)
来源仓库:https://github.com/andresnator/agents-orchestrator
仓库路径:skills/rfc
安装命令:
npx skills add https://github.com/andresnator/agents-orchestrator --skill rfc
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。

skills.shnpx skills
npx skills add https://github.com/andresnator/agents-orchestrator --skill rfc

这个 Skill 是做什么的

开发类 Skill 面向写代码、改代码和排查问题,一般会读写仓库文件,部分还会执行构建或测试命令。

rfc 用于处理 GitHub 仓库、Issue 和 Pull Request 协作信息,适合围绕代码变更进行整理。适用于开发流程中的仓库状态跟踪和协作事项管理。可通过 npx 命令从 agents-orchestrator 仓库安装,建议查看原始 README 了解具体能力。

使用前需确认权限范围和维护状态,避免误操作私有仓库或敏感提交。输出信息应以公开协作数据为基础,不擅自推断内部设计或未发布功能。

来源仓库内路径:skills/rfc

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 https://github.com/andresnator/agents-orchestrator --skill rfc;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

    它需要需要联网,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

rfc 怎么安装?

在终端执行 npx skills add https://github.com/andresnator/agents-orchestrator --skill rfc。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

rfc 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。

rfc 需要什么权限?

权限级别是需要联网。该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

rfc 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

rfc 用的人多吗?

公开数据是累计安装 238 次,近一周安装 10 次,下载 83 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。

rfc 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

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。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 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。
  • 需要处理开发类任务、又不想每次都手写提示词的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

开发类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
skill-vetter

技能审查员 · OpenClaw

8,779,018986操作浏览器MIT-0通过
ontology

本体论 · OpenClaw

6,636,960576只读MIT-0通过
entra-app-registration

内部应用程序注册 · Codex、Claude

6,631,872764接触敏感数据未明确通过
entra-app-registration

内部应用程序注册 · Codex、Claude

3,757,680201接触敏感数据未明确通过
nano-banana-pro

纳米香蕉专业版 · OpenClaw

3,526,898364接触敏感数据MIT-0有提醒
rfc

射频卡 · 本页

238—需要联网未明确通过

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

RFC Creator — Interactive Interview & Markdown Generator

Overview

This skill creates professional RFC (Request for Comments) documents by guiding the user through a structured interview. Instead of asking the user to fill in a template, the agent conducts a conversational interview, collecting information piece by piece, and then assembles a polished Markdown RFC at the end.

The goal is to make writing an RFC feel like a conversation, not paperwork.

Interview Flow

The interview has 7 phases. Walk the user through them one at a time. After each phase, summarize what you captured and confirm before moving on. Use the ask_user_input tool when the question has clear bounded options; use open-ended prose questions for everything else.

Important behavioral notes:

  • Ask one phase at a time. Do not dump all questions at once.
  • After the user answers, briefly reflect back what you understood and confirm it before proceeding.
  • If the user gives a vague answer, ask a clarifying follow-up — but don't be annoying about it. One follow-up is enough; if they're still vague, work with what you have.
  • If the user says "skip" or "not sure yet" for any section, mark it as TBD in the final document and move on.
  • The user can say "go back" at any time to revise a previous section.

Phase 1: Identity & Metadata

Collect the basic metadata for the RFC header.

Ask (open-ended):

  • Feature name: "What's the name or short identifier for this proposal? (e.g., async-worker-pool, user-auth-v2SI)"
  • Author(s): "Who are the authors? (names or handles)"

Then ask (using ask_user_input):

  • Type: single_select from ["Feature", "Enhancement", "Bug Fix", "Deprecation", "Process Change", "Other"]
  • Status: single_select from ["Draft", "In Review", "Accepted", "Rejected", "Superseded"]

Auto-fill the start date with today's date.

Phase 2: Related Context

Ask (open-ended):

  • "Are there related components, services, or systems this affects?"
  • "Any related tickets, issues, or prior RFCs to reference? (JIRA, GitHub issues, links, etc.)"

These are optional — if the user has nothing, skip gracefully.

Phase 3: Summary

Ask (open-ended):

  • "Give me a one-paragraph summary of this proposal. What is it, at a high level?"

If the user gives a long explanation, distill it to a concise paragraph and confirm: "Here's how I'd summarize that — does this capture it?"

Phase 4: Motivation

Ask (open-ended):

  • "Why is this needed? What problem does it solve or what use case does it support?"
  • "What happens if we don't do this?"

These two answers get combined into the Motivation section.

Phase 5: Detailed Design

This is the meatiest section. Approach it conversationally:

  • "Walk me through the design. How would this work?"
  • Follow up with clarifying questions based on what they describe:

- "How does X interact with Y?" - "What happens in the edge case where...?" - "Can you give me a concrete example of how a user/developer would use this?"

If the user provides code snippets, diagrams (in text), or API shapes, include them in the final RFC as fenced code blocks.

Encourage specificity, but respect the user's level of detail. Not every RFC needs to be an implementation spec — some are directional.

Phase 6: Trade-offs & Alternatives

Ask (using ask_user_input for the first, open-ended for the rest):

  • Have you considered alternatives?: single_select ["Yes, I have specific alternatives", "I have some rough ideas", "No, I haven't thought about it yet"]

Then depending on their answer:

  • If they have alternatives: "What alternatives did you consider and why did you choose this approach over them?"
  • If rough ideas: "Tell me what you've got — even half-formed ideas are useful to capture."
  • If none: "That's fine. Let me ask it differently — what's the main drawback or risk of this proposal?"

Also ask:

  • "What are the drawbacks of this approach? Why might someone argue against it?"

Phase 7: Open Questions

Ask (open-ended):

  • "What parts of this design are still unresolved or need further discussion?"
  • "Is there anything you're specifically looking for feedback on?"

Generating the RFC

Once all phases are complete (or the user says "that's everything"), generate the RFC as a Markdown file.

Output Template

Read the template at references/template.md and use that exact structure for the generated .md file.

File Generation Rules

  1. Save the file as rfc-[feature-id].md where [feature-id] is the kebab-case version of the feature name.
  2. Write the file to /mnt/user-data/outputs/.
  3. Use present_files to share it with the user.
  4. After presenting, offer: "Want me to adjust anything? I can also add extra sections like Implementation Plan, Timeline, or Security Considerations if you need them."

Writing Quality

When assembling the RFC from the interview answers:

  • Smooth out the language so it reads as a cohesive document, not a Q&A transcript.
  • Expand terse answers into clear prose while preserving the user's intent.
  • Preserve any technical detail, code, or examples verbatim.
  • Use professional but approachable tone — an RFC should be clear and direct, not stuffy.
  • If the user marked anything as TBD, include a visible > **TBD**: [brief note] callout so reviewers know it's intentionally unresolved.

Optional Extended Sections

If the user requests them (or if they naturally come up during the interview), the RFC can include additional sections after "Unresolved Questions":

  • Implementation Plan: Phases, milestones, estimated effort.
  • Timeline: Target dates for review, implementation, rollout.
  • Security Considerations: Threat model, auth implications, data handling.
  • Testing Strategy: How this will be validated.
  • Rollback Plan: How to revert if things go wrong.
  • Migration Plan: Steps for transitioning from the current state.

Only include these if the user wants them — don't bloat the RFC.

平台分布

Codex

35.89%
按下载量换算30

Claude

29.9%
按下载量换算25

Cursor

20.02%
按下载量换算17

Gemini CLI

9.01%
按下载量换算7

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

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