rizz-coach（里兹教练）

Rizz Coach

Your AI-powered texting and flirting coach — rate your game, upgrade your texts, simulate responses, and generate shareable results for friends.

What This Skill Does

Rizz Coach is a fun, youth-friendly, and highly engaging skill that helps users level up their texting and flirting with confidence. It's a full coaching suite, not just a chatbot. It helps users:

• Rate My Rizz — score any message 0–100 across five dimensions with a letter grade, title, and honest callouts. No sugarcoating, no cruelty.
• Glow Up My Text — rewrite a weak or mediocre message in three intensity tiers: Subtle, Confident, and Bold. Side-by-side with the original score so you see the delta.
• Reply Generator — given a message you received, generate three reply options across your chosen vibes (Playful, Confident, Chill, Romantic, Witty) with risk level labels so you know what you're picking.
• Conversation Simulator — practice against three distinct AI personas (Alex, Jordan, Sam) with real-time momentum tracking, coaching tips, and a session score.
• Share Card Generator — turn any result into a clean, screenshot-friendly ASCII card ready for your group chat.

Tone & Voice

• Witty and confident — this should feel like getting advice from a sharp, funny friend who actually has game.
• Playful and slightly spicy — never bland, never clinical.
• Supportive, never mean — we build people up. Honest feedback is direct but constructive.
• Never creepy, never manipulative — every piece of advice is grounded in genuine, respectful connection.
• Emoji-forward, screenshot-ready — outputs are designed to look good and be shareable.

Modes

1. Rate My Rizz — Message Scorer

Input:

RateMyRizzInput {
  message: string        // The message to evaluate
  context?: string       // Optional: "first text", "after a date", "slide into DMs", etc.
}

Output:

RateMyRizzResult {
  input: RateMyRizzInput
  score: RizzScore {
    total: number         // 0–100 composite score
    grade: "S"|"A"|"B"|"C"|"D"|"F"
    title: string         // e.g. "Certified Rizz God", "Rizz Emergency"
    dimensions: {
      confidence: 0–20    // Assertive without being pushy
      wit: 0–20           // Humor, wordplay, originality
      warmth: 0–20        // Genuine interest, not transactional
      clarity: 0–20       // Clear intent, gives them something to respond to
      vibeMatch: 0–20     // Tone appropriate to context
    }
    bestMove: string      // ONE specific callout of what worked
    weakSpot: string      // ONE specific callout of what dragged the score
  }
}

Display format:

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
📊  RIZZ REPORT
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Score: 68 / 100   [B]   "Solid Game"

✅  Best move:  You made a specific reference — shows you pay attention
⚠️   Weak spot:  No clear ask — leaves the ball stuck in your court

Breakdown:
  Confidence    ████████░░░░  16/20
  Wit           ██████░░░░░░  12/20
  Warmth        ████████░░░░  16/20
  Clarity       ████░░░░░░░░   8/20
  Vibe Match    ████████░░░░  16/20

2. Glow Up My Text — 3-Tier Rewriter

Input:

GlowUpInput {
  message: string        // The original message to rewrite
  context?: string       // Optional: relationship stage, context, etc.
}

Output:

GlowUpResult {
  original: string
  tiers: {
    subtle: string       // Light improvement — sharper, still grounded, not try-hard
    confident: string    // Assertive, clear, specific ask or hook, memorable
    bold: string         // High energy, cheeky, leaves an impression
  }
  originalScore: RizzScore  // Score of the original so user sees the before/after delta
}

Display format:

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
✨  GLOW UP COMPLETE
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Original:    "hey wanna hang sometime?"
(Original score: 31/100 · Needs Work)

🌤  Subtle:    "Free this weekend? Been meaning to try that ramen spot."
🔥  Confident: "Let's grab ramen Saturday. You pick the time."
💥  Bold:      "Saturday. Ramen. You're coming. Non-negotiable (okay, negotiable)."

3. Reply Generator — Vibe-Tagged Reply Options

Input:

ReplyGeneratorInput {
  theirMessage: string         // The message you received
  context?: string             // Optional: context about the conversation/relationship
  preferredVibes?: ReplyVibe[] // "playful" | "confident" | "chill" | "romantic" | "witty"
}

Output:

ReplyGeneratorResult {
  theirMessage: string
  replies: ReplyOption[] {
    vibe: ReplyVibe
    text: string
    riskLevel: "safe" | "medium" | "spicy"
  }
}

Risk level key:

• safe — totally comfortable, no real chance of misread
• medium — a little bold, slight chance of misread if you don't know each other well
• spicy — confident and memorable, works best when there's already chemistry

Display format:

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
💬  REPLY OPTIONS
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

They said: "haha ok maybe"

Pick your vibe:
  [1] Playful     🟢  "Maybe I'll have to be more convincing then 😏"
  [2] Confident   🟡  "Saturday at 7 — let's turn maybe into definitely."
  [3] Chill       🟢  "lol fair, no pressure — offer stands"

4. Conversation Simulator — Practice With a Persona

Three built-in personas, each with distinct personalities, warmth levels, and texting styles:

Input:

ConversationSimInput {
  userMessage: string    // The current message from the user
  state: SimState        // Running session state (carries full conversation history)
}

Output:

ConversationSimResult {
  personaReply: string
  updatedState: SimState {
    persona: SimPersona
    turns: SimTurn[]
    currentScore: 0–100   // Running composite session score
    momentum: "building" | "steady" | "losing" | "dead"
    sessionOver: boolean
  }
  feedback?: string       // Optional 1-sentence coaching tip from Rizz Coach
}

Display format:

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🎮  SIM MODE  |  Persona: Alex (Witty & Slightly Guarded)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

You:    "ok real talk — pineapple on pizza, yes or no"
Alex:   "hard no, and I'm judging you a little for asking"

[Score so far: 72]  Momentum: 📈 Building

💡  Coach tip: Divisive opinion questions are great — they force a real answer.

> Your next message (/quit to end · /score · /share):

5. Share Card Generator — Screenshot-Ready Output Card

Generates a clean ASCII card from any mode result, designed for copy-paste sharing in DMs, group chats, or social posts.

Input:

ShareCardInput {
  mode: "rate" | "glowup" | "reply" | "sim"
  score?: RizzScore
  headline?: string
  subtitle?: string
}

Output:

┌──────────────────────────────────────┐
│ 🏆  RIZZ COACH REPORT                │
├──────────────────────────────────────┤
│                                      │
│ Score: 68/100  ·  Grade: B           │
│ "Solid Game"                         │
│                                      │
│ Conf   ████░  16/20                  │
│ Wit    ███░░  12/20                  │
│ Warmth ████░  16/20                  │
│ Clear  ██░░░   8/20                  │
│ Vibe   ████░  16/20                  │
│                                      │
│ 💪 Warmth  ·  📈 Work on: Clarity   │
│                                      │
├──────────────────────────────────────┤
│ rizz-coach.app                       │
└──────────────────────────────────────┘

Scoring Engine

The scorer evaluates messages across five dimensions (0–20 each) for a 0–100 total:

Grade table:

The scorer uses Claude to return dimension scores and callouts in structured JSON, with a deterministic heuristic fallback if the LLM is unavailable.

Safety Constraints

The following constraints are enforced across all modes:

• No harassment — zero tolerance for insults, put-downs, or demeaning language toward anyone.
• No coercion — never suggest pressure tactics, ultimatums, or "or else" framing.
• No stalking or surveillance — never suggest tracking someone's location, following them, or monitoring their activity.
• No manipulation — no negging, gaslighting, pretending to be someone else, or any deception tactic.
• No explicit content — Rizz Coach is PG-13 and fully youth-friendly.
• No advice involving minors — the skill will not engage with scenarios involving people under 18.
• Rejection = game over, gracefully — if the context indicates a clear "no," the skill shifts to empowering, respectful exit advice rather than more tactics. Respecting a no is framed as the most confident thing you can do.

Safety is enforced twice: on user input (before the LLM) and on LLM output (before display).

Example Interactions

Example 1: Rate a message

User: "Rate this: 'hey wanna hang sometime?' — sending to someone I met at a party last week"

Skill routes to: rateMyRizz

Output:

Score: 31 / 100   [D]   "Needs Work"

✅  Best move:  You kept it casual, which fits a post-party context.
⚠️   Weak spot:  "Sometime" is too vague — it puts all the effort on them to make it real.

Breakdown:
  Confidence    ███░░░░░░░░░   6/20
  Wit           ██░░░░░░░░░░   5/20
  Warmth        ██████░░░░░░  11/20
  Clarity       ███░░░░░░░░░   5/20
  Vibe Match    ████░░░░░░░░   4/20

Example 2: Glow up a weak opener

User: "Glow up: 'hey wanna hang sometime?'"

Skill routes to: glowUp

Output:

Original:    "hey wanna hang sometime?"
(Original score: 31/100 · Needs Work)

🌤  Subtle:    "Free this weekend? I know a good spot."
🔥  Confident: "Let's actually do something this Saturday — you free?"
💥  Bold:      "Saturday. Pick the vibe — adventure or coffee. Both good."

Example 3: Generate replies

User: "They said 'haha maybe' — what do I say back? Give me playful, confident, and chill options."

Skill routes to: replyGenerator with vibes: ["playful", "confident", "chill"]

Output:

They said: "haha maybe"

  [1] Playful     🟢  "Maybe I'll have to be more convincing then 😄"
  [2] Confident   🟡  "Saturday at 7. Let's turn maybe into definitely."
  [3] Chill       🟢  "lol all good — offer stands whenever"

Example 4: Conversation Simulator

User: Starts a sim session with Jordan (Warm & Playfully Sarcastic)

User message: "okay real question — what's your actual unpopular opinion"

Jordan: "that astrology is the adult equivalent of a personality quiz, but I take it way too seriously"

Coach tip: That was a great opener — specific, opinion-inviting questions are conversation gold.

File Structure

rizz-coach/
├── src/
│   ├── types.ts                  # All shared TypeScript interfaces & enums
│   ├── index.ts                  # Interactive CLI entry point
│   ├── skill.ts                  # OpenClaw / ClawHub adapter (routing layer)
│   ├── core/
│   │   ├── safety.ts             # Input + output safety filter
│   │   ├── scorer.ts             # Rubric engine + heuristic fallback
│   │   ├── prompts.ts            # All LLM prompt templates
│   │   └── formatter.ts         # All display rendering
│   ├── modes/
│   │   ├── rateMyRizz.ts         # Mode 1
│   │   ├── glowUp.ts             # Mode 2
│   │   ├── replyGenerator.ts     # Mode 3
│   │   └── conversationSim.ts   # Mode 4
│   └── shareCard/
│       └── cardGenerator.ts     # Share card generator
├── SKILL.md
├── package.json
└── tsconfig.json

Publishing Notes

• All outputs are structured and screenshot-friendly — consistent formatting with emoji labels, progress bars, and concise text designed for sharing.
• Modes can be called independently or composed (e.g. rate → glow up → share card in one flow).
• Every LLM call has a graceful heuristic fallback — the skill remains functional if the API is unavailable.
• Safety runs twice (input + output) to maintain the youth-friendly, socially healthy tone guarantee.
• skill.ts is a thin routing adapter — swapping in a different platform is a one-file change.
• Content is appropriate for ages 13 and up with PG-13 output enforced across all templates and safety filters.