Token导航 LogoToken导航

Agent Skill · 开发

build-talk-outline(建立谈话大纲)

build-talk-outline 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。

build-talk-outline 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 21,184。

总安装 194周安装 8GitHub Stars 4下载 63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:build-talk-outline(建立谈话大纲)
来源仓库:https://github.com/phazonoverload/devadvokit
仓库路径:skills/build-talk-outline
安装命令:
npx skills add https://github.com/phazonoverload/devadvokit --skill build-talk-outline
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/phazonoverload/devadvokit --skill build-talk-outline

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。

build-talk-outline 基于 DevRel 原则生成演讲大纲,需先读取用户本地配置文件获取背景信息。适用于技术分享会、会议发言等结构化内容策划,强调清晰逻辑与重点突出。必须引用 winston-talk-principles.md 中的结构性要求,不可随意省略。

首次使用需先运行 /setup-devadvokit 完成上下文设置,否则无法继续。适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

来源仓库内路径:skills/build-talk-outline

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/phazonoverload/devadvokit --skill build-talk-o…;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

build-talk-outline 怎么安装?

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

build-talk-outline 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。

build-talk-outline 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。

build-talk-outline 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。

build-talk-outline 用的人多吗?

公开数据是累计安装 194 次,近一周安装 8 次,下载 63 次,GitHub 4 星。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。

build-talk-outline 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
build-talk-outline

建立谈话大纲 · 本页

1944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

Before doing anything else:

  1. Check if ~/.devadvokit.md exists.
  2. If it does, read it silently and use it throughout this skill.
  3. If it does not, stop and tell the user: "I need your DevRel context before I can run this skill. Please run /setup-devadvokit first."

Read reference/winston-talk-principles.md silently before starting. Apply these principles throughout — they are not optional polish, they are structural requirements.


Q&A

Ask these questions one at a time. Wait for each answer before moving on.

  1. Paste your CFP abstract or talk description. If you don't have one yet, describe the talk in a few sentences — the problem it addresses, the angle you're taking, and what you want the audience to leave with.
  2. What's the session format and length? (e.g. talk / 30 min, workshop / 90 min — include whether Q&A is expected and if so, how long it typically runs at this event)
  3. Any specifics about this particular audience beyond your general context? (e.g. expected seniority mix, whether this is a specialist or general-track event, anything you know about what they've already seen at this conference)
  4. Where does this talk sit on the theory-to-practical spectrum?

- Theory-heavy — concepts, mental models, arguments. Code or tooling appears as illustration, not instruction. - Balanced — roughly equal time on ideas and hands-on material. - Practical-heavy — mostly code, tooling, or process. Concepts are introduced only as needed to support the practical work.

  1. Are you planning any live demos? If yes: describe each one — what you're showing, what it depends on (network, auth, third-party services, local tooling), and roughly how long you'd expect each to take.
  2. Are there any audience interaction moments planned? (e.g. polls, show-of-hands questions, short exercises, pair discussions) If so, describe them.
  3. Is there anything that absolutely cannot be cut if you run short on time — a specific point, demo, or moment that the talk doesn't work without?
  4. How do you want to end? (e.g. clear call to action, open question for the audience, resources slide, straight into Q&A)

Building the outline

Use the CFP material, the Q&A answers, and the speaker context from ~/.devadvokit.md to construct the outline. Before generating, consult reference/outline-structure-patterns.md to select the most appropriate structural pattern for the talk type and format.

Timing rules

  • Reserve Q&A time first if the user indicated it. A standard allowance is 5 min for a 30-min slot, 10 min for a 45–60 min slot, unless the user specified otherwise.
  • The opening section is always 2–3 minutes. The close is always 2 minutes. These are fixed.
  • For live demos, add a 20% time buffer on top of the user's estimate. Demos always run longer than expected.
  • Flag if the total content the user described won't fit the format. Suggest what to cut or defer rather than silently compressing everything.

Section format

Present the outline as a structured sequence. For each section:

[START TIME–END TIME] Section title
Content notes: what happens in this section, key points covered
Type: [concept / demo / interaction / transition / close]

Use cumulative timestamps from 0:00. Be precise — "5:00–9:00" not "about 5 minutes in".


Winston requirements — apply to every outline

Read reference/winston-talk-principles.md and apply the following before finalising the outline. These are structural checks, not suggestions.

Opening (Winston)

The opening section must contain an empowerment promise — a specific statement of what the audience will be able to do by the end that they can't do now. Write it out in the content notes for the opening section. It must be specific enough that a person in the audience could evaluate at the end whether the promise was kept.

Flag and remove any of the following if the user's described opening contains them:

  • Opening with a joke
  • "Thank you for having me" or any variant
  • Apologies (for anything)
  • Extended self-introduction before the promise is made

The Winston Star — memorable moment

Before writing the outline, identify where the following land. Include them as content notes in the relevant sections:

  • Slogan — the short phrase that becomes the handle for the talk's core idea. If one doesn't emerge naturally from the CFP material, note where in the outline it should be introduced and suggest a candidate.
  • Surprise — the counterintuitive point. The assumption the audience holds that this talk challenges. Name it in the content notes for the section where it lands.
  • Salient idea — the single thing. State it explicitly in the outline. If the CFP material implies more than one salient idea, flag this: the talk will be harder to remember.
  • Symbol — a concrete visual or physical object that represents the core idea. Suggest one based on the talk content. If nothing physical fits, note the closest verbal equivalent.

Structure check (Winston)

If the talk is persuasive (making a case, recommending a change, arguing a position):

  • Vision must be established within the first 5 minutes. Note this in the outline.
  • Every section must either advance the vision or provide proof of work. Flag any section that does neither — it's a cut candidate.

Close (Winston)

The close section must end with a contributions slide — what the audience now has that they didn't have when they walked in. Write out what that slide should contain in the content notes for the close.

The close must mirror the opening promise. Check that the empowerment promise made at the start is directly answered by the contributions at the end.

Flag if the user described ending with "Questions?" or "Thank you" as the final slide.

Slide density check (Winston)

Based on the number of sections and the content in each, estimate whether the slide count is likely to exceed the Winston threshold (20 slides for a 30-min talk is the warning level). Flag it if so and note which sections are most likely to produce dense slides.


Demo risk assessment

For each live demo identified in the Q&A, produce a risk block immediately after the section it appears in. Consult reference/demo-risk-patterns.md to assess each demo's risk profile.

Format each risk block as:

⚠️ Demo risk: [demo name]
Risk level: [Low / Medium / High / Critical]
Dependencies: [list what this demo requires to work]
Failure modes: [what is most likely to go wrong]
Fallback: [specific fallback for this demo — see below]
Prep checklist: [what to do before the talk to reduce risk]

Fallback hierarchy — apply in order:

  1. Pre-recorded video — record the demo running successfully the day before. Keep it under the demo's time budget. Play it if live fails.
  2. Screenshots with narration — a deck of screenshots showing the key steps and outputs, talked through live.
  3. Code walkthrough — open the finished code or config in an editor and walk through it statically, explaining what it would do.
  4. Audience imagination — only for low-complexity demos. Describe the expected output verbally and move on. Never use this for demos where the visual result is the point.

Winston's demo rule: Note in every demo's risk block what the speaker should say if the demo fails mid-run. The failure should teach something — prepare a sentence that frames it as illustrative rather than broken.

State explicitly which fallback applies and what the speaker needs to prepare in advance to have it ready.


Risk summary

After the full outline, produce a consolidated risk section:

Overall risk level: [Low / Medium / High] — based on the number and type of live demos, the proportion of the talk that depends on them, and how much of the content can survive if they fail.

Critical path: identify which section, if it goes wrong, most damages the talk. This is usually the demo or moment the user said they can't cut.

Time risk: flag if the outline is tight. A talk with no slack is a talk that runs over. Recommend where to trim if needed — be specific about which section and by how much.

Contingency plan: a one-paragraph description of what the speaker should do if things start going wrong — too slow, demo fails, audience engagement drops. Make it concrete and actionable, not generic advice.


Output order

Present in this order:

  1. Structured outline with timestamps and Winston annotations inline
  2. Demo risk blocks (inline, after each relevant section)
  3. Risk summary
  4. Offer: "Want me to adjust the balance, restructure a section, or build out speaker notes for any part of this?"

Before presenting any output, read ../../shared/ai-antipatterns.md and silently rewrite any flagged patterns. Do not mention this step to the user.

平台分布

Codex

34.92%
按下载量换算22

Claude

29.64%
按下载量换算19

Cursor

19.76%
按下载量换算12

Gemini CLI

10.22%
按下载量换算6

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