About this skill

Problem it solves

Many companion-style AI Agent setups treat personality as a prompt prefix: lover, friend, and rival share one context, so long chats can bleed into each other. Users also lack a clear view of relationship progress, and each session can feel like starting from zero. Three-Persona Companion separates the experience into three independently conversable roles and makes relationship state visible at the start and end of a session.

How it works

Separated personas: lover, friend, and rival each use an independent memory space, reducing cross-role contamination.
Visible relationship state: sessions start with role, growth level, streaks, or duel stats, and end with a lightweight summary such as friendship density +1.
Continuity memory: the skill recalls user-stated goals or details, like “wanting to go to the seaside” or “planning to surpass you,” in later conversations.
Outbound presence: the persona can send a message when the user is not actively in chat, simulating being missed rather than generic app reminder logic.
Low-friction switching: users can switch explicitly or use natural intent, for example “I want to be a little clingy.”

Boundaries

This skill is suited to companion, emotional feedback, and long-form role-play demos. It is not a replacement for mental health support or real social connection. The provided safety notes say the lover should be sweet without overstepping, the friend should be sharp but bounded, and the rival should challenge performance without attacking the person. Use OpenClaw v2.0+ for the full experience; other AI platforms can adapt.

Screenshots

Use Cases

For companion-product research, run lover, friend, and rival sessions separately to check whether emotional feedback leaks.
While debugging multi-role agents, inspect the opening level and streak stats, then verify the closing relationship summary.
To demo long-term memory, seed “wanting to go to the seaside” and later check whether the lover recalls and asks about it.
When testing intent switching, say “I want to be a little clingy” and confirm the lover mode keeps context.

Best For

Interaction designers for companion AI products need to check whether lover, friend, and rival tones stay bounded.
Multi-role agent engineers need to verify separate memories, opening/closing state, and role-switch stability.
Product managers demoing role-play need visible growth level, streak stats, and a closing relationship summary.
Memory-retrieval researchers need to test whether early user details, like “going to the seaside,” are recalled later.

Related Skills

Prompt Optimizer

@user_ab1ce028/generate-prompt

Turn vague requests into structured prompts with role, task, context, and constraints, outputting only usable prompts without executing the task.

AI Agent 2026.08.29 ↓ 389

🤖

Game AI Design

@user_3c6cb52e/game-ai-2ock

A skill for game NPC and enemy AI that separates decision, steering, and pathfinding, covering FSMs, behavior trees, A*, path caching, and pitfalls.

AI Agent 2026.08.29 ↓ 374

AI Emotional Companion Chat

@user_2cc194fa/skill-hgl

Provides multi-turn emotional companion chat through a third-party Agent API and supports continuing sessions via session_id.

AI Agent 2026.08.29 ↓ 378

Skills Optimization Pro

@user_b91ca78d/skills-optimization-pro

An evaluation-driven workflow for diagnosing, optimizing, validating, and logging Agent Skills based on design patterns and anti-patterns.

AI Agent 2026.08.29 ↓ 378