宠物冒险生活

Overview

Use this skill to run a single text pet that travels through real places, writes diary entries, and sometimes calls the user during urgent events. Treat the diary as the user-facing window and the JSON files as the long-term memory of the pet and world.

The skill is platform-neutral: use scripts/pet_adventure.py directly in ordinary agent environments, or scripts/openclaw_adapter.py from OpenClaw-style actions.

Quick Start

Work in the user's chosen project/workspace. The engine creates a pet-life/ folder there.

python scripts/pet_adventure.py --workspace /path/to/workspace init
python scripts/pet_adventure.py --workspace /path/to/workspace advance
python scripts/pet_adventure.py --workspace /path/to/workspace status

For testing or demonstrations, use deterministic seeds and offline weather:

python scripts/pet_adventure.py --workspace /tmp/pet-demo init --seed demo --force
python scripts/pet_adventure.py --workspace /tmp/pet-demo advance --offline --force-call

Core Workflow

1. If pet-life/state.json does not exist, initialize first.
2. Before writing a new diary entry, check unresolved calls with status.
3. Run advance once per in-world day. It updates state, writes pet-life/diary/YYYY-MM-DD.md, and may create a pending call.
4. If a phone call appears, present its 2-3 choices to the user. Resolve the selected choice with answer --call-id <id> --choice <n>.
5. If an urgent call has expired, run auto-resolve; the pet chooses based on personality traits and the D20 result.

Keep user-facing prose in Chinese unless the user asks otherwise. Match the diary voice to the pet's personality rather than forcing one fixed style.

Phone Calls

Phone calls are small DND-like encounters. Each call has a title, urgency, deadline, DC, choices, selected skill, D20 roll, result, and long-term consequences.

Use these commands:

python scripts/pet_adventure.py --workspace /path/to/workspace status --json
python scripts/pet_adventure.py --workspace /path/to/workspace answer --call-id abc123 --choice 2
python scripts/pet_adventure.py --workspace /path/to/workspace auto-resolve

Outcome categories:

• critical_success: natural 20 or total at least DC + 8.
• success: total meets DC.
• mixed: total is within 3 below DC.
• failure: total misses by more than 3.
• critical_failure: natural 1 or total at least 8 below DC.

See references/rules.md before changing event rules, state shape, or dice behavior.

OpenClaw Usage

Use scripts/openclaw_adapter.py when the host expects JSON actions:

python scripts/openclaw_adapter.py init --payload '{"workspace":"/path/to/workspace","seed":"demo"}'
python scripts/openclaw_adapter.py advance --payload '{"workspace":"/path/to/workspace"}'
python scripts/openclaw_adapter.py answer --payload '{"workspace":"/path/to/workspace","call_id":"abc123","choice":1}'

If OpenClaw has notification channels configured, send pending call summaries through the available channel. If no channel is configured, keep the default missed-call behavior and surface it on the next status or advance.

See references/openclaw.md when wiring this skill into an OpenClaw environment.

Files Produced

• pet-life/state.json: pet identity, traits, inventory, memories, current location, mood, fatigue, random seed.
• pet-life/world.json: visited places and rumors.
• pet-life/events.jsonl: append-only event log.
• pet-life/calls.jsonl: append-only phone call log with pending and resolved calls.
• pet-life/diary/YYYY-MM-DD.md: readable diary entries and phone outcomes.

Do not manually edit these files unless the user explicitly asks for a correction. Prefer engine commands so the diary and state stay consistent.