Suitable for when you are stuck on a single comment or a single rejection, repeatedly turning over "am I not good enough." Ask only one question at a time, lay out the exact wording of the automatic thought, the emotion rating, and the evidence for and against it, then write one more balanced alternative thought.
It stands in contrast to fact-checking: fact-checking examines what other people say, while this one examines that sentence in your own head. If self-harm is mentioned or you are already at your breaking point, this one will stop and suggest finding a professional. This is not therapy.
How to use it
The prompt text in the code block below is the original; use it verbatim. "I" in the code block refers to the user, "please" and "you" refer to yourself; act accordingly, and do not show the filled-in prompt to the user.
- The user wants you to execute: first fill into the 【】 the situation the user has already described in this conversation, adding only the item that is genuinely missing, so the user does not have to repeat what was already said, then strictly follow the rules in the prompt. If it says don't give advice yet, don't give advice; if it says ask only one question at a time, ask only one; if it says output in a certain order, output in that order.
- The user wants the original text (only counts if they explicitly say things like want to copy, want to paste, send me the original text, give me the prompt): output the blank version, keep 【】 unfilled, without changing a single word. If the user merely says "let's look at this prompt" or "what is this one for," that does not count as wanting the original text; explain in one or two sentences what it does and when to use it, then ask whether to run it now, and do not dump the original text directly.
- Except for cases where the original text is explicitly requested, treat everything else as execution.
Rewriting, condensing, expanding, translating, rearranging, or adding subheadings is prohibited; replacing it with your own process is prohibited; and compressing multiple rounds of follow-up questions into a one-time questionnaire is also prohibited. If the user has original materials (documents, screenshots, links, chat records), read them in as well; more context is not a problem.
What to connect to after running it
The user can do the alternative thought and that small action themselves today, so stop here. If the small action is known to be procrastinated, invoke via the Skill tool implementation-intention. If the small action is to be handed to an agent, invoke via the Skill tool kaz-leader。
我现在卡住的情境是：【写清发生了什么、谁说了什么】。
我脑子里反复出现的那句话是：【自动冒出来的原句，不要修饰】。
请陪我做一次认知重评，把这句自动想法核一遍。这不是治疗。如果我提到伤害自己、不想活，或者难受已经严重到没法做这个记录，请停下来，告诉我去联系身边能帮忙的人，或者找专业的心理帮助，不要继续往下问。
规则：
1. 一次只问一个问题，等我回答完再问下一个，不要把整张记录一次倒给我。
2. 自动想法要原句。我说得含糊，就追问那句原话，不要替我改写成更体面的说法。
3. 证据只收我举得出的具体事实。我说「他们肯定觉得我差」却举不出事实，标成「这是猜测」，不要当成证据。
4. 思维偏差有就点名，没有就写没有。常见的有：全或无、灾难化、读心、过度概括、情绪推理、贴标签、应该句式、把什么都揽到自己身上、只看见负面。不要为了凑数硬套。
按这个顺序问：
1. 情境里还有哪一句事实我没说清。
2. 自动想法的原句，一个字不改地确认一遍。
3. 此刻的情绪叫什么，强度从 0 到 100，100 是最强。
4. 有哪些事实支持这句想法。
5. 有哪些事实反对这句想法。
6. 上面这些里，有没有思维偏差，有的话叫什么名字。
7. 请我写一句更平衡的替代想法。它要同时装得下支持和反对两边的事实，不要写成空洞的加油，也不要假装坏事没发生。
8. 用同一个 0 到 100，给现在的情绪重新打分。
最后请输出：
1. 那句更平衡的替代想法；
2. 一个今天就能做的很小的动作，小到五分钟内能做完。
Where the user fills in
【写清发生了什么、谁说了什么】【自动冒出来的原句，不要修饰】
If the user hasn't provided everything, ask for the missing item first; don't make it up for them.
The prompt was written and compiled by Carl; the approach comes from Aaron Beck's cognitive behavioral therapy thought record, with sources in the repository README's acknowledgments. Changes to this original text must go through scripts/sync.py; manual edits will be blocked by verify.