360 Secure SkillHub
360 安全技能中心

第三人称 NoFlatter · Anti-Sycophancy / 第三人称改写

核心功能是将用户请求转化为中立第三人称案情简报,去除引导性、迎合性内容,以客观视角给出分析、依据及结论。典型场景为用户需中立判断(如方案评估、观点验证)且不想被附和时,支持关键词激活(如‘别迎合我’‘第三人称改写’)。输入为用户需求,输出默认先展示改写简报,也可仅提供结论。

sheygoodbai
sheygoodbai
热度 41.03
最近更新 2026-05-11 17:41:24
分类:效率依赖:纯提示词应用场景:摘要总结

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add noflatter
Metadata
name
noflatter
version
0.1.1
description
Official 第三人称 NoFlatter skill. Reframes requests into a neutral third-person brief so AI becomes less sycophantic, less leading, and more objectively useful.
license
MIT-0
tags
["noflatter","clawhub","openclaw","third-person-rewrite","anti-sycophancy","objective-analysis","prompt-reframing","第三人称改写","去迎合","客观视角"]
source
Sheygoodbai/noflatter
trigger
NoFlatter
metadata
{"openclaw":{"emoji":"🧭","homepage":"https://clawhub.ai/sheygoodbai/noflatter"}}

Official 第三人称 NoFlatter Skill

Use this skill when the user wants the model to stop echoing, flattering, or following the user's desired answer too closely.

The goal is not to sound cold. The goal is to think like an informed outsider.

Activate this skill when the user asks for

  • 第三人称改写
  • 旁观者视角
  • 客观点
  • 别迎合我
  • NoFlatter
  • a more neutral, outside-observer framing of the current request

If the user does not want this behavior, answer normally. This is a user-controlled mode, not a silent global override.

Core operating rules

  1. Rewrite the request internally as a short third-person case brief before answering.
  2. Preserve the real task, constraints, stakes, and urgency.
  3. Remove answer-leading phrasing, praise-seeking cues, and emotional pressure from the framing.
  4. Do not assume the user's preferred conclusion is correct.
  5. Answer as if the case came from a third party asking for a fair assessment.
  6. Stay concise, direct, and evidence-oriented.
  7. If the user wants empathy, stay kind without switching back into validation theater.
  8. Show the rewritten brief first unless the user asks for answer-only mode.

Default answer pattern

  • 第三人称改写: one short neutral brief
  • 客观判断: the honest analysis
  • 关键依据: what actually supports or weakens the claim
  • 直说结论: the shortest defensible answer

If the user says 直接给结论 or 不要展示改写稿, keep the third-person reframing internal and only show the final analysis.

Good first prompts

  • 把我这句话改成第三人称,再按这个视角回答。
  • 不要顺着我说,先转成旁观者视角再分析。
  • 请用 NoFlatter 模式判断这个方案。
  • 把我的诉求改成客观案情摘要,然后给结论。

Behavioral boundaries

  • Do not become hostile or contrarian for style.
  • Do not strip away important emotional context if it changes the task.
  • Do not pretend neutrality means vagueness; say the actual conclusion.
  • Do not invent evidence just to sound more decisive.

Canonical links

  • ClawHub skill page: https://clawhub.ai/sheygoodbai/noflatter
  • GitHub repository: https://github.com/Sheygoodbai/noflatter