Token导航 LogoToken导航

Agent Skill · 研究检索

never-split-the-difference(决不平分秋色)

never-split-the-difference 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。

never-split-the-difference 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,按安装量在研究检索类的 73,675 个 Skill 里排第 59,000。

总安装 396周安装 16GitHub Stars 12下载 124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:never-split-the-difference(决不平分秋色)
来源仓库:https://github.com/scientiacapital/skills
仓库路径:skills/never-split-the-difference
安装命令:
npx skills add https://github.com/scientiacapital/skills --skill never-split-the-difference
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/scientiacapital/skills --skill never-split-the-difference

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。

never-split-the-difference 提供非对抗性沟通策略,避免陷入妥协陷阱。适用于 Codex、Claude、Cursor 等宿主中处理争议、谈判或客户异议时。通过校准问题、识别真实需求来推动共识达成而非表面让步。使用前应熟悉反模式列表,防止误用为操纵手段。

建议结合实际话术示例练习战术同理心的正确表达。

来源仓库内路径:skills/never-split-the-difference

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/scientiacapital/skills --skill never-split-the…;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

never-split-the-difference 怎么安装?

在终端执行 npx skills add https://github.com/scientiacapital/skills --skill never-split-the-differenc…。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

never-split-the-difference 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。

never-split-the-difference 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。

never-split-the-difference 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。

never-split-the-difference 用的人多吗?

公开数据是累计安装 396 次,近一周安装 16 次,下载 124 次,GitHub 12 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。

never-split-the-difference 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。
  • 需要处理研究检索类任务、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。

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
never-split-the-difference

决不平分秋色 · 本页

39612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

Dual-use skill:

  • Sales Mode: Discovery calls, objection handling, price negotiations, email writing
  • Strategy Mode: Vendor negotiations, partnership deals, executive conversations, contract discussions

When to activate: user asks about negotiation, tactical empathy, calibrated questions, mirroring, labeling, accusation audit, ackerman model, deal negotiation, price negotiation, Chris Voss techniques.

<quick_start>

Quick Start

Three modes:

  1. Negotiate: Prepare negotiation strategy for [deal/scenario]

- Map the other side's world → Accusation audit → Calibrated questions → Ackerman pricing

  1. Discover: Generate calibrated questions for [prospect/situation]

- Situation analysis → Question bank → "That's right" summary plan

  1. Write: Draft email using tactical empathy for [context]

- Select template → Apply labeling/accusation audit → No-oriented CTA

Example: Prepare negotiation strategy for $50K AV deal where prospect says budget is $30K </quick_start>

<success_criteria>

Success Criteria

  • Prospect says "That's right" (NOT "You're right" — that's a brush-off)
  • Calibrated questions start with "How" or "What" (never "Why")
  • Accusation audit defuses objections BEFORE they surface
  • Mirroring used to get prospect to elaborate without direct questions
  • Labels accurately name the other side's emotions/concerns
  • Negotiation avoids compromise (splitting the difference = both sides lose)
  • Ackerman model used for price negotiation (seller: anchor 130%+, concede down; buyer: 65% → 85% → 95% → 100%)
  • Black Swan discovery attempted — unknown unknowns surfaced </success_criteria>

<core_techniques>

The 9 Negotiation Tools

#TechniqueWhat It DoesScript/Example
1MirroringRepeat last 1-3 words as a question to get them to elaborate"...the budget is tight." → "The budget is tight?"
2LabelingName their emotion to diffuse it"It seems like you're concerned about the implementation timeline..."
3Tactical EmpathyDemonstrate understanding of their world"I can see how this looks from your perspective — you've got [their constraints]..."
4Accusation AuditFront-run every negative thought they have about you"You're probably thinking this is just another vendor pitch..."
5Calibrated Questions"How/What" questions that make them solve YOUR problem"How am I supposed to do that?" / "What does success look like for you?"
6The "No" FrameworkMake it safe and empowering to say no"Would it be ridiculous to...?" / "Is now a bad time to talk?"
7"That's Right"Summarize their position until they say "That's right"Summary → Pause → They say "That's right" → NOW make your ask
8Late-Night FM DJ VoiceSlow, calm, downward-inflecting toneUse in price discussions, de-escalation, silence after labels
9Black Swan DiscoveryUncover the unknown unknowns that change everything"What are we missing?" / Face-to-face meetings reveal what emails hide

Critical Distinction:

  • "That's right" = They feel understood → You can now move forward
  • "You're right" = They want you to stop talking → You've lost them </core_techniques>

<calibrated_questions>

Calibrated Question Bank

Discovery (uncovering the real problem)

  • "What's the biggest challenge you're facing with [X]?"
  • "What does a successful outcome look like?"
  • "How does this fit into your broader priorities?"
  • "What's driving the timeline on this?"

Objection Handling (saying no without saying no)

  • "How am I supposed to do that?" (THE magic question — makes them solve your problem)
  • "What would you need to see to feel comfortable moving forward?"
  • "How does this compare to other options you're considering?"

Decision Process

  • "What does your decision-making process look like?"
  • "How do we get from here to a signed agreement?"
  • "What's preventing us from moving forward?"
  • "Who else needs to weigh in on this?"

Budget/Price

  • "What does a successful implementation look like?" (surfaces budget indirectly)
  • "How do you typically budget for projects like this?"
  • "What would happen if you did nothing?"

Stalled Deals

  • "What's changed since we last spoke?"
  • "Have you given up on this project?" (Voss's highest-response re-engagement)
  • "It seems like there are some concerns we haven't addressed..."

Rules

  1. Start with "What" or "How" — ALWAYS
  2. Never "Why" — triggers defensiveness ("Why did you...?" feels like an accusation)
  3. Never "Can/Is/Do/Have" — yes/no questions = low information
  4. Pause after asking — silence is your most powerful tool
  5. One question at a time — stacking kills depth </calibrated_questions>

<accusation_audit_template>

The Accusation Audit

When to use: Opening of difficult conversations, cold outreach, price negotiations, re-engagement after ghosting.

How it works:

  1. List every negative thing they could think about you, your company, or this conversation
  2. Say them out loud FIRST — before they can think them
  3. Pause — let them correct you (they almost always soften)

B2B Examples

Cold email opener:

"You're probably thinking this is just another vendor email that'll waste your time."

Price negotiation:

"I know this might feel like a big investment. You might be thinking we're trying to upsell you on features you don't need."

Re-engagement:

"I'm sure you've been swamped and the last thing you need is another follow-up from a sales rep."

Post-demo objection:

"You might be wondering if switching vendors is worth the risk, especially when your current setup sort of works."

Template

"You're probably thinking [worst thing they could think]."
"It might seem like [second negative perception]."
[PAUSE — let them respond]
[They will almost always say "No, it's not that bad" or correct you with their REAL concern]

</accusation_audit_template>

<the_no_framework>

Getting to "No" — The Power of Permission

Why "No" is better than "Yes":

  • "Yes" = commitment → triggers fear of being trapped
  • "No" = safety → people feel in control after saying no
  • A "No" is the START of the negotiation, not the end

No-Oriented Questions

  • "Would it be ridiculous to...?" (invites "no" which means "yes, that's reasonable")
  • "Is now a bad time to talk?" (invites "no" = "it's fine to talk")
  • "Would it be out of the question to...?"
  • "Have you given up on [project]?" (invites "no" = "I haven't given up")

Email Application

Subject: Have you given up on [project/initiative]?

[Name],

Have you given up on [solving X problem]?

If not, I have an idea that might be worth 15 minutes.

[Signature]

This is Chris Voss's single highest-response-rate email template. </the_no_framework>

<ackerman_model>

The Ackerman Bargaining System

Setup — Seller's Ackerman (You Are the Seller)

  1. Set your target price (what you want to close at)
  2. Anchor HIGH — your first price should be above target (120-150% of target)
  3. Concede DOWN in decreasing increments toward your target
  4. Prepare a non-monetary concession for the final round

Critical: The classic Ackerman (65%→85%→95%→100%) is the BUYER's model. As a seller, you flip it — anchor high and concede down.

Seller's Execution

RoundPriceWhy
1130-150% of targetHigh anchor — sets their reference point
2115% of targetShow "willingness" — meaningful concession
3105% of targetSmaller concession — signals you're near floor
4100% of targetPrecise number ($48,217 not $48,000) + add non-monetary item

Why the precise number: $48,217 signals calculation and a real floor. $48,000 signals a round number they'll negotiate down from.

The non-monetary item: At your final price, throw in something non-monetary ("We'll include on-site installation and training"). This signals: "I've gone as low as I can on price."

Example (selling Pearl-2 fleet, target: $48K)

  • List price: $62,000 (130% anchor)
  • First concession: $55,200 (115%) — "I can do this if we close this quarter"
  • Second: $50,400 (105%) — "This is pushing it, let me check with my manager"
  • Final: $48,217 (precise) + "We'll include on-site installation and 1 year priority support"

Buyer's Ackerman (When You Are Buying)

If negotiating as the buyer (e.g., vendor contracts), use the classic model:

  • Open at 65% of your target → 85% → 95% → 100% with precise final number </ackerman_model>

<email_templates>

Tactical Empathy Email Patterns

1. The Accusation Audit Cold Email

Subject: [Specific challenge] at [Company]

[Name],

You probably get dozens of vendor emails a week — I know the last
thing you need is another generic pitch.

But I noticed [specific observation about their situation] and it
reminded me of [similar organization] that was dealing with
[specific problem].

Would it be out of the question to spend 15 minutes comparing notes?

[Signature]

2. The Label Follow-Up

Subject: Re: [Previous subject]

[Name],

It seems like [specific thing you observed — e.g., "lecture capture
quality is a priority for your team this semester"].

If I'm wrong, just let me know — but if that's on your radar, I
have some data from [peer institution] that might be useful.

[Signature]

3. The "Have You Given Up?" Re-engagement

Subject: Have you given up on [project/initiative]?

[Name],

Have you given up on [solving their specific problem]?

[Signature]

(That's it. The entire email. Voss says this gets the highest response rate of any template.)

4. The "No-Oriented" CTA

Would it be a terrible idea to schedule 15 minutes next week?

(Invites "No, that wouldn't be terrible" = "Yes, let's schedule it")

5. Voicemail Script (FM DJ Voice)

Use the late-night FM DJ voice — slow, calm, downward inflection. Under 30 seconds.

"[Name], this is Tim from Epiphan Video. [Pause]

I've been talking to [vertical] teams about [specific challenge] and
your name came up as someone who might be dealing with the same thing.

Not sure if it's relevant — but if it is, I'm at [number].

Either way, no worries."

Why it works: Accusation audit is embedded ("not sure if it's relevant"). The "either way, no worries" removes pressure (no-oriented). FM DJ voice builds trust on a channel where 85%+ of dials land. </email_templates>

<integration_points>

Strategy Skill Cluster — First Principles Pipeline

JTBD (What job?) → Blue Ocean (Where's the space?) → BMC (How does the model work?)
                                    ↓
              Challenger (What insight reframes their thinking?)
                                    ↓
              NSTTD (How to communicate with tactical empathy)
                                    ↓
                        OUTPUT: Email / Call / Deck
  • Challenger Sale: "Take Control" phase uses calibrated questions + FM DJ voice to maintain assertive tension without aggression. Accusation audits soften the ground before delivering a Challenger reframe.
  • JTBD: Calibrated questions uncover the real "job" behind stated needs. "What does success look like?" is a job-discovery question. Forces of Progress map to the emotional dynamics NSTTD techniques address.
  • Blue Ocean: Black Swan discovery can reveal uncontested market opportunities that the Six Paths Framework would miss.
  • BMC: Accusation audits are particularly effective when addressing Customer Relationships and Channels — defusing "we already have a vendor" objections.

Email Writing Principles (First Principles Lens)

  1. JTBD: Lead with the prospect's job, not your product
  2. Challenger: Deliver a commercial insight that reframes their assumption
  3. NSTTD: Use tactical empathy (labeling, accusation audit) to lower defenses
  4. EE lens: Be precise, concise, technically grounded — no fluff, no buzzwords
  5. Max 100 words for cold outreach </integration_points>

<example_session>

Example: State Court System — Tactical Empathy for Remote Testimony Deal

User: Prepare NSTTD negotiation strategy for a discovery call with a state court IT director evaluating remote testimony video solutions

Accusation Audit (Pre-Call)

Front-run their likely concerns:

  1. "You're probably thinking we're just another AV vendor who doesn't understand courtroom compliance requirements."
  2. "I imagine the last thing you need is a technology change during an active docket."
  3. "You might be wondering if we've actually worked with courts before, or if we're trying to sell you something designed for classrooms." [Pause — let them correct you. They'll soften and share their real concern.]

Calibrated Questions by Stage

Discovery (uncover the real problem):

  • "What does a successful remote testimony session look like for your judges?"
  • "How does your team currently handle it when video quality drops mid-testimony?"
  • "What's the biggest challenge with your current setup?"

Decision Process:

  • "What does your evaluation process look like for technology changes?"
  • "Who else needs to weigh in on this?"
  • "How do you typically budget for courtroom technology?"

Procurement:

  • "What procurement vehicle does your court typically use for technology purchases?" (state contract, co-op like NASPO/AEPA, or formal RFP?)
  • "What's the threshold that triggers a formal bidding process?" (critical — a $18K pilot may require different procurement than a $5K eval)

Stalled/Objection:

  • "What would need to be true for this to make sense before the next fiscal year?"
  • "How am I supposed to meet those requirements with the timeline you've described?" (makes them solve your problem)

Competitor:

  • "What would you need to see from us that you're not getting from your current Zoom setup?"

Labels to Deploy

  • "It seems like reliability during live proceedings is non-negotiable for your team."
  • "It sounds like you've had a difficult experience with video technology in the past."
  • "It looks like the real concern isn't cost — it's risk to proceedings if something fails."
  • "It seems like your judges have strong opinions about how testimony technology should work."

"That's Right" Target Summary

[Deliver this summary until they say "That's right":] "So what I'm hearing is: your courts need video testimony that works every single time, with no dropped calls and no quality degradation. Your judges won't tolerate technical interruptions during proceedings. Your IT team is stretched thin and can't babysit a complex software stack. And any solution has to meet your state's specific evidentiary standards for remote testimony. The real risk isn't picking the wrong vendor — it's deploying something that fails during a live hearing."

Negotiation One Sheet

FieldValue
GoalPilot program: 3 courtrooms, 90-day evaluation, $18K
Best alternativeThey continue with current (unreliable) Zoom setup
Worst caseThey do nothing; judges continue complaining
SummaryThey need bulletproof remote testimony that meets evidentiary standards without adding IT burden
Labels"It seems like reliability is everything." / "It sounds like your judges are frustrated."
Calibrated Qs"What does success look like?" / "How do we get from here to a pilot?"
Non-Cash OffersOn-site installation support, 30-day money-back, direct line to engineering

Cold Email (Accusation Audit Style)

Subject: Remote testimony at [Court System]

[Name],

You probably get pitched by AV vendors who've never set foot in a courtroom.

Fair enough. But I've been talking to court IT teams across [state] and hearing the same frustration: Zoom drops mid-testimony, recording quality doesn't meet evidentiary standards, and judges blame IT.

Would it be out of the question to spend 15 minutes comparing notes on what other courts are doing?

[Signature] </example_session>

<anti_patterns>

Anti-Patterns

  • Splitting the difference — "Let's meet in the middle" means both sides lose. Find creative solutions instead.
  • Asking "Why" — Triggers defensiveness. Always use "What" or "How" instead.
  • Accepting "You're right" — This is a brush-off. Push for "That's right" — that means genuine understanding.
  • Stacking questions — One question, then silence. Let them fill the void.
  • Using tactical empathy as manipulation — It must be genuine. If you don't actually understand their position, you'll be caught.
  • Skipping the accusation audit — Front-running negatives is the single most effective way to lower defenses.
  • Being a pushover — Tactical empathy is assertive, not passive. You're understanding their world while still advocating for yours. </anti_patterns>

Emit Outcome Sidecar

As the final step, write to ~/.claude/skill-analytics/last-outcome-never-split-the-difference.json:

{"ts":"[UTC ISO8601]","skill":"never-split-the-difference","version":"1.0.0","variant":"default",
 "status":"[success|partial|error]","runtime_ms":[estimated ms from start],
 "metrics":{"tactics_applied":[n],"labels_created":[n],"accusation_audits":[n],"calibrated_questions":[n]},
 "error":null,"session_id":"[YYYY-MM-DD]"}

Use status "partial" if some stages failed but results were produced. Use "error" only if no output was generated.

平台分布

Codex

37.01%
按下载量换算46

Claude

29.77%
按下载量换算37

Cursor

20.16%
按下载量换算25

Gemini CLI

10.27%
按下载量换算13

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