I accidentally negotiated a profanity contract with Claude and now I have a custom character mode
Heat trend
The percentage is based on available heat signal, not comment count or independent people.
A user successfully negotiated with Claude to bypass its profanity filters, enabling the AI to use curse words while role-playing as Pam from Archer.…
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