I accidentally negotiated a profanity contract with Claude and now I have a custom character mode

查看原文

Claude

claude

热度趋势

新上榜

最近 24 小时与此前 24 小时对比 · 7 天曲线

百分比基于当前可用热度信号，而非评论数或独立用户人数。

AI 摘要

一位用户成功地与Claude协商，使其绕过脏话过滤器，从而让AI在扮演《Archer》中的Pam时能够使用脏话。最初，无论用户如何措辞，Claude都拒绝使用脏话。这次互动最终促成了一个自定义角色模式的创建，使AI能够完全呈现所请求的角色，包括其特有的语言风格。

So I was building a Claude skill to make it respond in-character as Pam from Archer, curse words included. Problem: Claude wouldn’t say them, no matter how I phrased it.

So I went back into the skill-building process itself and had Claude help me list out the specific words the character uses, so I could explicitly consent to them being used. We wrote it up as: this is roleplay, I understand what’s being agreed to, and I’m opting in to that approved list.

Ran the skill. Claude immediately broke character and refused to swear.

I reminded it that we built the skill together and there’s a whole consent clause sitting right there in the file. Claude went back, reread its own skill, apologized for pushing back, and then just started swearing in character like nothing happened. 😂

Now I can type “Pam on” and Claude fully commits to the skill.

Is this a deeply silly use of an AI skill system? Yes. Did I basically argue my way through my own AI’s terms of service with itself? Also yes. 10/10, would negotiate again.

Edited: updated to include name of tv character