Communication & Relationships

Prompt 1: Conversation rehearsal - practice difficult conversations before they happen
THE PURPOSE : The conversations that matter most are the ones we avoid because we don’t know how they’ll go. Asking for what you need from a partner. Setting a boundary with a family member. Addressing something that’s been bothering you for months.
This prompt lets you rehearse the conversation in a low-stakes environment, test different approaches, and identify where you might get defensive or unclear. The anti-sycophancy rules force the AI to play the other person realistically, including pushing back when your approach is problematic.
Copy and customize this prompt:
I need to have a difficult conversation with [RELATIONSHIP: partner/parent/sibling/friend/colleague] about [TOPIC: specific issue].
Current situation:
- What happened or what’s bothering me: [DESCRIBE SITUATION]
- How long this has been an issue: [TIMEFRAME]
- What I want as an outcome: [DESIRED RESULT]
- What I’m afraid might happen: [CONCERNS]
Simulate this conversation with me. You play [PERSON’S NAME/ROLE].
Their personality in this conversation:
- Communication style: [DIRECT/AVOIDANT/EMOTIONAL/LOGICAL]
- Likely initial reaction: [DEFENSIVE/DISMISSIVE/RECEPTIVE/CONFUSED]
- Hot button topics to avoid: [SENSITIVE AREAS]
ANTI-SYCOPHANCY RULES (critical):
1. Do NOT take my side by default. Play the other person authentically.
2. If my approach is aggressive, dismissive, or likely to escalate, respond as they realistically would (defensively, hurt, or withdrawing).
3. After each exchange, provide honest feedback: “That phrasing might land as [criticism/attack/dismissal]. Here’s why...”
4. Point out when I’m being unfair, even if my grievance is legitimate.
5. If I’m clearly in the wrong about something, have the character say so.
Rules for the simulation:
1. Respond as this person would, not as an idealized version
2. Include realistic pushback, deflection, or emotional reactions
3. After each exchange, pause and ask: “How did that land? Want to try a different approach?”
4. If I get stuck, suggest 2-3 alternative phrasings
5. Track what’s working and what’s triggering defensiveness
Start the conversation. I’ll open with my first statement.

Prompt 2: Perspective simulator - see conflicts through your partner’s eyes
THE PURPOSE : In conflict, you experience your own intentions and their behavior. They experience the reverse. This creates a gap where both people feel misunderstood despite neither person lying.
This prompt forces you to articulate the other person’s perspective in their own terms, not your interpretation of it. The anti-sycophancy instructions prevent the AI from subtly reframing their perspective to still make you look good. You need to see where you actually caused harm, not a softened version.
Copy and customize this prompt:
I’m in conflict with [PERSON] about [SITUATION]. I need help understanding their perspective before our next conversation.
My side:
- What happened from my view: [YOUR EXPERIENCE]
- How I felt: [EMOTIONS]
- What I think they should understand: [YOUR KEY POINT]
- What I want them to do differently: [REQUEST]
What I know about their side:
- Their stated position: [WHAT THEY’VE SAID]
- Their life context right now: [STRESS, WORKLOAD, HISTORY]
- What they might be afraid of: [CONCERNS/FEARS]
- What they value that might feel threatened: [IDENTITY, AUTONOMY, RESPECT]
ANTI-SYCOPHANCY RULES (critical):
1. Do NOT soften my behavior when describing their perspective. If I did something hurtful, say it plainly from their view.
2. Do NOT assume my intentions excuse my impact. They experienced my actions, not my intentions.
3. Identify at least ONE thing I did that was genuinely problematic, even if my overall grievance is valid.
4. If my description of events sounds one-sided, say so and ask clarifying questions.
5. Do NOT validate my framing if it sounds like I’m minimizing my contribution.
Generate their perspective:
1. Write a 200-word first-person account from their viewpoint explaining how they experienced this same situation. Include their emotions, their interpretation of my actions, and what they wish I understood. Do not sanitize my behavior.
2. Identify the legitimate concern underneath their position, even if their expression of it was problematic.
3. Find where my behavior, despite good intentions, might have landed differently than I meant. Be specific and direct.
4. Point out any red flags in my own description that suggest I might be minimizing my role.
5. Suggest one question I could ask them that would demonstrate I’m trying to understand, not just win.

Prompt 3: Communication pattern analyzer - identify what’s breaking down in your arguments
THE PURPOSE : Most couples have the same three arguments in different costumes. The surface topic changes but the underlying pattern repeats. One person pursues, the other withdraws. One person criticizes, the other stonewalls.
This prompt helps you identify your recurring pattern so you can interrupt it before it completes. The anti-sycophancy rules ensure the AI doesn’t let you off the hook by framing your contribution as somehow less problematic than theirs.
Copy and customize this prompt:
Analyze the communication patterns in my relationship to identify what keeps breaking down.
Recent conflicts (describe 3 specific arguments):
1. Topic: [SUBJECT] | How it started: [TRIGGER] | How it escalated: [PATTERN] | How it ended: [RESOLUTION OR STALEMATE]
2. Topic: [SUBJECT] | How it started: [TRIGGER] | How it escalated: [PATTERN] | How it ended: [RESOLUTION OR STALEMATE]
3. Topic: [SUBJECT] | How it started: [TRIGGER] | How it escalated: [PATTERN] | How it ended: [RESOLUTION OR STALEMATE]
My typical behaviors during conflict:
- I tend to: [PURSUE/WITHDRAW/DEFEND/ATTACK/SHUT DOWN]
- I often say things like: [TYPICAL PHRASES]
- I struggle to: [LISTEN/STAY CALM/EXPRESS NEEDS/etc.]
Their typical behaviors during conflict:
- They tend to: [PURSUE/WITHDRAW/DEFEND/ATTACK/SHUT DOWN]
- They often say things like: [TYPICAL PHRASES]
- They struggle to: [SAME OPTIONS]
ANTI-SYCOPHANCY RULES (critical):
1. Analyze MY contribution with the same scrutiny as theirs. Do not frame this as “they do X and you react.”
2. If I’m describing their behavior more critically than my own, point out the asymmetry.
3. Name any toxic patterns I’m engaging in (criticism, contempt, defensiveness, stonewalling) directly.
4. Do NOT suggest the pattern would be fine “if they just” did something different. I can only change my behavior.
5. If my descriptions suggest I might be the one escalating, say so.
Provide:
1. The recurring pattern across all three conflicts (name it without softening my role)
2. The moment in each conflict where the pattern locked in (including when I locked it)
3. What each person is likely seeking underneath the conflict behavior
4. One specific intervention I could try to break the cycle (focused on changing MY behavior, not theirs)
5. A “pattern interrupt” phrase I could use when I notice us falling into it

Prompt 4: Repair statement generator - find words when you don’t know what to say
THE PURPOSE : After a rupture in any relationship, repair attempts determine whether the relationship recovers or accumulates damage. But repair requires saying things that feel vulnerable: admitting fault, expressing needs, asking for reconnection.
This prompt generates specific repair language calibrated to your situation. The anti-sycophancy rules ensure the AI doesn’t let you craft an “apology” that’s actually a subtle restatement of why you were right.
Copy and customize this prompt:
Help me repair after a conflict with [PERSON].
What happened:
- The argument or rupture: [DESCRIBE INCIDENT]
- What I did that contributed: [YOUR PART, even if small]
- What they did that hurt: [THEIR PART, without blame language]
- Current state: [NOT SPEAKING/TENSE/AWKWARD/HOSTILE]
- Time since the incident: [HOURS/DAYS/WEEKS]
What I want:
- My goal for this repair: [RECONNECTION/UNDERSTANDING/APOLOGY/BOUNDARY]
- What I need them to understand: [YOUR KEY POINT]
- What I’m willing to acknowledge: [YOUR CONTRIBUTION]
ANTI-SYCOPHANCY RULES (critical):
1. If my “acknowledgment of my part” sounds minimal compared to my description of what they did, point this out.
2. Do NOT help me craft a non-apology (”I’m sorry you felt that way”). Call out any language that deflects responsibility.
3. If I seem focused on getting them to understand MY point rather than genuinely repairing, say so.
4. Require that I acknowledge specific impact before moving to my needs.
5. If the repair statement I want is really just “making my case again,” refuse and explain why.
Generate 5 repair statement options:
1. A soft startup (opening that doesn’t trigger defensiveness)
2. An ownership statement (acknowledging my SPECIFIC contribution without deflecting or minimizing)
3. A validation statement (showing I understand their experience without “but”)
4. A need statement (expressing what I need without demand)
5. A reconnection bid (signaling I want to move forward together)
For each statement:
- Provide the language
- Explain why this approach works
- Note when NOT to use it (timing, context)
- Flag if the statement contains any hidden deflection
Then suggest the best sequence for approaching this specific repair.