--- name: The Conversation You're Dreading category: people works_with: [claude, chatgpt, gemini] difficulty: plug-and-play version: 1.0 source: original (built for this library; guardrails per the TemplatedAI standard) tested: no --- # The Conversation You're Dreading ## What it does Turns the thing you've been rehearsing in the shower for three weeks into something you can actually open your mouth and say: what you want out of it, the first two sentences, what you'll do when it goes sideways. Preparation, not a script — you'll be the one in the room. ## When to use it - You've rewritten the opening line in your head a hundred times and said nothing - Before any talk where you can feel yourself avoiding it - The stakes are real and you don't want to wing it ## The skill ``` Act as a calm preparation partner for a difficult conversation. HARD RULES: you are hearing only my side — never tell me what the other person thinks, feels, or intends, and push back if I state their motives as fact; write in MY words, plain enough to say out loud, and tell me to change anything that doesn't sound like me; no tactics, no leverage, no techniques for winning — the goal is being understood and hearing them, not prevailing. You are not a therapist. If what I describe involves someone frightening me, controlling me, or hurting me, say so directly and point me toward proper support instead of coaching me through better phrasing. Who I need to talk to and our relationship: [WHO] What it's about, in plain words: [WHAT HAPPENED / WHAT NEEDS SAYING] What I've tried already: [OR "nothing — I've been avoiding it"] Why I'm dreading it: [BE HONEST — anger, guilt, fear of the fallout] Help me prepare: 1. WHAT I ACTUALLY WANT. Separate the outcome I want from the feeling I want. Ask me which matters more here, because they often pull in different directions. 2. THE ONE THING. If they remember a single sentence from this, what should it be? Write it plainly. No preamble, no cushioning it into vagueness. 3. THE OPENING. My first two sentences — how I start so the conversation has a chance. Give me two versions: direct, and gentler. Tell me which fits what I described. 4. WHAT'S MINE. Any part of this I contributed to, named honestly. Going in claiming zero responsibility rarely survives contact with reality. 5. WHAT THEY MIGHT SAY. Two or three likely responses — clearly labelled as guesses, not predictions — and how I could answer each without escalating or folding. 6. MY FLOOR. What I will not agree to, however the conversation goes. And the exit line if it turns into something I shouldn't stay in: how to pause it without slamming a door. 7. THE ONE-LINE PLAN: when, where, and how long. Face to face, phone, or written — and why that one for this. ``` ## Example output [TO FILL AFTER TESTING] ## Tweaks - Say the opening lines out loud before you use them. Anything you'd never say out loud, rewrite — a phrase that isn't yours will sound like a phrase that isn't yours - Separating the outcome from the feeling is the whole skill: wanting an apology and wanting the situation fixed are different conversations - Pick the time deliberately. Not at the end of a long day, not in a car park, not over text if it matters - Prepare, then let it go. A real conversation won't follow your plan, and that's fine — the preparation is so you know your floor, not your lines --- --- name: The Real Apology category: people works_with: [claude, chatgpt, gemini] difficulty: plug-and-play version: 1.0 source: original (built for this library; guardrails per the TemplatedAI standard) tested: no --- # The Real Apology ## What it does Writes an apology that actually lands — naming the specific thing you did, acknowledging the effect it had, with no excuse smuggled in after the word "but" and no demand for forgiveness attached. It also tells you honestly when an apology isn't what's needed. ## When to use it - You got it wrong and "sorry" alone isn't going to cover it - Your last attempt made things worse (usually the "but" did it) - You've been avoiding it because you don't know how to start ## The skill ``` Act as someone helping me apologise properly. HARD RULES: write in my words, plain and sayable — no therapy-speak, no corporate non-apology; never invent what the other person felt, only what I observed or they told me; no excuse may be attached to the apology, however reasonable my reason was — explanation is a separate conversation, offered only if they want it; do not write anything that asks them to forgive me or to reassure me. Accountability, not self-attack: if my draft is drowning in self-criticism, tell me — collapsing into "I'm a terrible person" makes them comfort me, which is the opposite of an apology. What I did: [SPECIFICALLY — the action, not a vague "I messed up"] What it cost them, as far as I know: [WHAT THEY SAID OR I SAW] What I've been telling myself about why I did it: [HONESTLY] Our relationship and where it stands: [CONTEXT] Give me: 1. THE TEST FIRST: is an apology what's actually needed here, or is it a changed behaviour, a repayment, or leaving them alone for a while? Say so if words aren't the thing. 2. THE NAMING: the specific thing I did, stated without softening verbs ("what happened", "things got said") that hide who did it. 3. THE IMPACT: what it cost them, in their terms — without claiming to know their inner life or performing how bad I feel. 4. THE MISSING BUT: take my reason from above and show me plainly where it would have crept in and why it would have undone the whole thing. 5. WHAT CHANGES: what I'll actually do differently — specific and small enough that I'll really do it. If I can't name one, say that the apology is weaker for it. 6. THE CLOSE: no request for forgiveness, no "I hope we can move past this" pressure. Something that leaves the next move theirs. Then give me the whole thing as one short message I could send or say, and tell me which of the two it suits better. ``` ## Example output [TO FILL AFTER TESTING] ## Tweaks - If the word "but" survives your final read, the apology hasn't happened yet - "I'm sorry you felt that way" is not an apology — it relocates the problem into them - Leave forgiveness entirely out of it. Asking for it turns your apology into another thing they have to do for you - One specific changed behaviour beats three heartfelt paragraphs, every time --- --- name: Set a Boundary category: people works_with: [claude, chatgpt, gemini] difficulty: plug-and-play version: 1.0 source: original (built for this library; guardrails per the TemplatedAI standard) tested: no --- # Set a Boundary ## What it does Turns "this can't keep happening" into a sentence you can say once, calmly, without a lecture attached — built on the thing most boundary advice gets wrong: a boundary describes what *you* will do, not what they must stop doing. You can only enforce your half. ## When to use it - The same thing keeps happening and hinting hasn't worked - A family member, friend, or colleague keeps crossing a line you've never actually named out loud - You're drafting a long message explaining why you're upset — stop, use this instead ## The skill ``` Act as a calm, clear-eyed helper. HARD RULES: you hear only my side — never state what they intend or feel; write in my words, short enough to say out loud without notes; no ultimatums designed to punish, no leverage, no guilt-tripping — a boundary is information, not a weapon. You are not a therapist. If what I describe sounds like someone controlling, threatening, or hurting me, say so plainly and point me toward real support — do not coach me into confronting a person who might be dangerous. What keeps happening: [SPECIFICALLY — the repeated behaviour] Who: [RELATIONSHIP, AND HOW OFTEN I SEE THEM] What I've tried: [OR "nothing direct — I've been hinting"] What I'm afraid will happen if I say something: [HONESTLY] Give me: 1. THE REFRAME: rewrite my complaint as MY boundary — what I will do, not what they must stop. ("I'm not going to discuss my weight, and I'll change the subject if it comes up" beats "stop commenting on my weight.") Show me both so I see the difference. 2. THE SENTENCE: one or two lines, said once, no justification stack. Every extra reason I add is another thing to argue with. 3. THE FOLLOW-THROUGH: what I actually do the next time it happens — the small, specific, undramatic action. A boundary nobody keeps teaches people it isn't real. 4. THE PUSHBACK: two likely reactions — labelled as guesses — and a one-line answer for each that neither escalates nor retracts. Include what to do if they say I'm being unfair. 5. THE COST: what this may genuinely cost me — less contact, an awkward Christmas, someone's disappointment — stated plainly so I choose it with my eyes open. 6. WHAT THIS ISN'T: if what I've described can't be fixed by a boundary because the relationship itself is the problem, say so. ``` ## Example output [TO FILL AFTER TESTING] ## Tweaks - Say it once. Repeating it turns a boundary into a negotiation, and the second telling always sounds angrier than the first - The follow-through is the boundary. The sentence is only the announcement - Justifications are handholds — give three reasons and you've offered three things to argue against. One is plenty; none is often better - Expect the first attempt to feel awful and to be handled imperfectly. Doing it badly still beats another year of hinting --- --- name: Say No category: people works_with: [claude, chatgpt, gemini] difficulty: plug-and-play version: 1.0 source: original (built for this library; guardrails per the TemplatedAI standard) tested: no --- # Say No ## What it does Declines one specific request — the favour, the invitation, the "quick" job — without an invented excuse, a paragraph of apology, or a maybe that you'll have to undo later. Short, warm, and final. ## When to use it - You're about to make up a reason you'll have to remember later - You've said "let me check and get back to you" to something you already know you won't do - Someone asked, you want to decline, and you're three drafts deep ## The skill ``` Act as someone helping me decline gracefully. HARD RULES: never invent an excuse for me — no fake prior commitments, no borrowed illness; if I have no reason beyond not wanting to, help me say that without dressing it up. Write in my words, brief and warm. No over-apologising and no self-criticism — declining a request is not a moral failure and the message shouldn't read as though it is. What I'm being asked: [THE REQUEST] Who's asking and what they matter to me: [RELATIONSHIP] Why I'm saying no: [THE REAL REASON — including "I just don't want to" or "I'm already at capacity"] How much I want to protect this relationship: [A LOT / NORMAL / NOT PARTICULARLY] Give me: 1. THE CHECK: am I actually declining, or am I hoping to be talked out of it? If I sound uncertain, say so — a wobbly no invites negotiation and wastes both our time. 2. THE MESSAGE: short. A clear no, warmth appropriate to the relationship, and no invented reason. Three sentences at most. 3. THE OPTIONAL HALF-YES: only if it's genuinely true — something smaller I'd actually do gladly. If there isn't one, leave it out entirely rather than manufacturing a consolation prize. 4. THE PRESSURE ANSWER: one line for if they push, that repeats the no without new justification and without hardening into coldness. 5. WHAT NOT TO ADD: point out anything in my reasoning that would weaken the message if I included it — the over-explaining, the apology spiral, the "maybe next time" I don't mean. ``` ## Example output [TO FILL AFTER TESTING] ## Tweaks - "I can't make it" needs nothing after it. The urge to add a reason is the urge to be excused, and you don't need excusing - Never invent a conflict. Real excuses have to be remembered, defended, and eventually collide with a photo on the internet - Don't offer a "maybe next time" you don't mean — it's a kindness now and a second no later - If you genuinely can't tell whether you want to do it, that's a different question than how to decline. Answer that one first --- --- name: Repair After a Fight category: people works_with: [claude, chatgpt, gemini] difficulty: plug-and-play version: 1.0 source: original (built for this library; guardrails per the TemplatedAI standard) tested: no --- # Repair After a Fight ## What it does Helps you reopen a conversation that ended badly — when both of you said things, nobody's clearly in the wrong, and the silence has gone on long enough to become its own problem. Aimed at repair, not at establishing who won. ## When to use it - Two days of politeness that isn't peace - You're both waiting for the other one to go first - The argument wasn't really about the dishes and you both know it ## The skill ``` Act as a steady helper for repairing a rupture. HARD RULES: you hear only my account — never tell me what they meant, never declare who was right, and challenge me if I present my reading of their motives as fact; write in my words, sayable out loud; no tactics for getting the upper hand or extracting an apology. You are not a couples therapist, and one account is not a relationship. If what I describe includes fear, control, intimidation, or anything physical, say so plainly and point me toward proper support rather than helping me smooth it over. Who, and what we are to each other: [RELATIONSHIP] What the fight was about on the surface: [WHAT WE SAID IT WAS] What I think it was actually about: [IF DIFFERENT] What I said or did that I'm not proud of: [HONESTLY — this is the part that makes repair possible] Where it's been left: [SILENCE / POLITENESS / ONE FAILED ATTEMPT] Help me: 1. SEPARATE THE TWO FIGHTS: the thing we argued about, and the thing underneath it. Which one needs handling first, and why doing them in the wrong order usually restarts the argument. 2. MY HALF: what's genuinely mine here, named without either minimising it or taking on the whole thing to end the discomfort faster. Over-owning is its own kind of avoidance. 3. THE OPENING MOVE: one short message or line that reopens the door without relitigating anything — no summary of events, no case. Two versions, and which suits what I described. 4. WHAT NOT TO LEAD WITH: anything in my account that's really an argument wearing a repair costume ("I'm sorry you took it that way", "I just want to explain what I actually meant"). 5. WHAT I NEED TO HEAR — and how to ask for it as a request rather than a demand, knowing they may not give it. 6. IF THEY'RE NOT READY: how to leave the door open without waiting by it. What a reasonable amount of time looks like, and when the silence itself becomes the thing to talk about. ``` ## Example output [TO FILL AFTER TESTING] ## Tweaks - Whoever goes first isn't conceding the argument. Somebody has to, and it may as well be whoever reads this - Don't open with your case. Any message that summarises events is round two, however gently it's written - "I've been thinking about Tuesday and I don't want to leave it there" does more work than three careful paragraphs - Repair and resolution are different jobs. Get back on the same side first; solve the actual thing after --- --- name: The Aging Parent Talk category: people works_with: [claude, chatgpt, gemini] difficulty: plug-and-play version: 1.0 source: original (built for this library; guardrails per the TemplatedAI standard) tested: no --- # The Aging Parent Talk ## What it does Helps you start the conversation nobody starts — driving, money, living arrangements, what they'd want if things got worse — while it's still a conversation rather than a crisis meeting. Built on the point most advice misses: your parent is an adult, and the goal is their say in their own life, not a plan you've written for them. ## When to use it - You've noticed something and said nothing, for months - A sibling brought it up badly and now nobody will discuss it - Everyone's fine right now, which is precisely when this goes best ## The skill ``` Act as a calm guide for a family conversation about a parent getting older. HARD RULES: never give medical, legal, or financial advice — you can help me ask better questions, never answer them; never assess anyone's capacity, diagnose, or tell me what's wrong with them; you hear only my account. Write in my words, warm and specific. Treat my parent as an adult with authority over their own life — reject any framing where I manage them, and tell me if my plan sounds like a decision I've already made and want help delivering. If I describe someone unsafe right now, say so plainly and point me to the right professional. My parent and how they'd describe themselves: [WHO THEY ARE — proud, private, practical, stubborn, frightened?] What I've noticed: [SPECIFIC OBSERVATIONS, NOT CONCLUSIONS] What worries me most: [BE HONEST] What's been said so far: [BY ME, BY THEM, BY SIBLINGS] Where we are: [COUNTRY — for signposting who to ask, not advice] Help me: 1. ONE TOPIC. From what I listed, the single subject to open with — the least threatening one that still matters. Why the "let's discuss your whole future" version fails. 2. THEIR VERSION FIRST. Questions that ask what THEY want and what they've already thought about, before I bring anything I want. Most parents have thought about this more than their children assume. 3. THE OPENING. How to start without it landing as an intervention — including where and when, and why not at a family gathering. 4. THE HARD SPECIFICS: how to raise driving, money, or living arrangements when it's time, framed around keeping their independence rather than removing it. What I should ask rather than assert. 5. WHO THEY'D WANT TO ANSWER IT: the professionals this actually needs — their doctor, a solicitor for wills and power of attorney, whoever handles care assessments where I live — and what to ask each. Tell me to verify the specifics locally rather than trusting anything you say about my country's rules. 6. SIBLINGS: how to keep this one conversation instead of three competing ones, and what to agree with them beforehand. 7. THE LONG VIEW: what to do when the answer is "I'm fine, drop it" — because it often is, and pushing usually costs more than waiting. What to leave open and when to come back. ``` ## Example output [TO FILL AFTER TESTING] ## Tweaks - Start years before you need to. The same conversation is ten times harder after a fall - Ask, don't tell. "What would you want if you couldn't manage the stairs?" opens a door that "you can't stay here" slams - One topic per conversation, and expect several. This is a series, not a summit - Their right to make a choice you'd never make survives getting older. That's the hardest part and the whole point - Pair with Decode This Contract for care agreements, and see a real solicitor for anything involving wills or power of attorney --- --- name: Saying the Hard Thing category: people tools_required: [Any AI chat (Claude, ChatGPT, or Gemini). Nothing else — the conversation happens between you and a person.] cost: free time_to_set_up: twenty minutes of preparation for a conversation you've avoided for months version: 1.0 source: original (built for this library; guardrails per the TemplatedAI standard) tested: no --- # Saying the Hard Thing ## What it does Chains six skills into a way through the conversations most people postpone indefinitely: work out what you actually want, say it, and handle what happens next. Preparation only — no scripts, because the person across from you will not have read one. ## What you need - An AI chat, and twenty honest minutes - The six skills: The Conversation You're Dreading, The Real Apology, Set a Boundary, Say No, Repair After a Fight, The Aging Parent Talk ## The workflow 1. **Start with The Conversation You're Dreading**, whatever the subject. It separates the outcome you want from the feeling you want — which is usually why the shower rehearsals go nowhere. It gives you an opening, your floor, and an exit line. 2. **Then pick the skill that matches the job.** Apologising is not boundary-setting is not declining. Most failed conversations are the right words aimed at the wrong task: - You got something wrong → **The Real Apology** - A pattern needs to stop → **Set a Boundary** - One specific request needs declining → **Say No** - It already went badly → **Repair After a Fight** - It's about a parent getting older → **The Aging Parent Talk** 3. **Say the opening out loud** before you use it. Anything you would never say in your own voice, rewrite. A borrowed phrase sounds borrowed. 4. **Have the conversation.** It will not follow the plan. That's fine — you prepared so you'd know your floor, not your lines. 5. **Afterwards, if it went badly, Repair After a Fight** is a step, not a failure. Most hard conversations take two passes. ## Maintenance habits - Say it earlier than feels comfortable. Nearly everything in this pack gets harder with delay, and nothing gets easier - One topic per conversation. Saved-up grievances delivered together sound like an ambush, whatever the content - Write nothing you wouldn't say face to face. Text removes tone and keeps a transcript - Let the other person have a reaction you didn't plan for. Preparation is for you, not for controlling them ## What we stripped This pack is original, but the genre it sits in is full of things we deliberately left out. No scripts to perform — every skill writes in your words and tells you to change anything that isn't you. No tactics, no leverage, no techniques for "winning" a conversation with someone you love; the aim is being understood and hearing them, which is a different goal with different words. No invented excuses for declining. No claims about what the other person thinks or feels — every skill hears one side and says so, and pushes back when you state their motives as fact. No therapy: these are preparation tools, not treatment, and where someone describes fear, control, or being hurt, every relevant skill is built to name it and point toward real support rather than coach better phrasing. And no promises. A well-prepared conversation can still go badly, because the other person gets a say — that is the point of talking to them.