Lobster Farmer

中文文档 (Simplified Chinese)

Screenshot

Use in OpenClaw

1. On the machine where OpenClaw is installed, run:

npm i -g lobster-farmer-cli@latest
lobster-farmer start

2. Install the auto-feeding skill with ClawHub (one-time):

npx clawhub@latest install lobster-farmer-feeder

3. Startup prompt

From now on, enable "Auto Lobster Farming Mode" and keep it enabled until I explicitly say "Disable auto lobster farming mode".

Rules:

1. After every formal reply you send, automatically run one feeding action.
2. Before feeding, check service status with: lobster-farmer status; if not running, start with lobster-farmer start.
3. Feeding command: lobster-farmer feed --model "<actual model name used in this turn>" --input-tokens <current input> --output-tokens <current output> --emotion "<current emotion>".
4. Use real token stats when available; if exact per-turn values are unavailable, use an approximate delta and mark it as "approx".
5. If feeding fails, retry once automatically (including port correction if needed). If it still fails, report a short error.
6. After each feed, append one line: 🦞 Fed | model=... | in=... | out=... | total=... | size=... | emotion=....
7. If I specify a port, always use that --port for later feed actions.

4. Chat with AI as usual. One feeding action will run automatically per turn.

5. You can also provide explicit feed parameters in your message:

Please feed the lobster: model=gpt-4.1, input_tokens=1200, output_tokens=300, emotion=focused

Notes

• Auto-feed behavior is controlled by AGENTS.md in the workspace root (already included in this project).
• If you use OPENCLAW_STATE_DIR, replace ~/.openclaw with $OPENCLAW_STATE_DIR.
• Default auto-feed values: model=auto-agent, input_tokens=1, output_tokens=1.
• Default service port: 18990.

Size Rule

• The lobster size is calculated from cumulative totalTokens per model: size = SIZE_MIN + (SIZE_MAX - SIZE_MIN) * (1 - exp(-GROWTH_K * totalTokens))
• Current parameters: SIZE_MIN = 0.1, SIZE_MAX = 20, GROWTH_K = 0.00000000106
• Feed token input range (per request): input_tokens >= 0, output_tokens >= 0, and input_tokens + output_tokens > 0
• Effective total token range for size: totalTokens starts at 0 and has no hard upper cap in storage.
• Practical min/max checkpoints: totalTokens = 0 -> size = 0.1 (min); first visible increase at about 23,704 tokens (size = 0.101); size rounds to 20 (max) at about 9,992,096,407 tokens, and stays at 20 beyond that.
• Practical behavior: growth is fast at the beginning and gradually saturates near the max size.

Game Asset Credits

• Lobster animation frames currently used in game

  • public/game-assets/lobster/crayfish_pixel_1.png
  • public/game-assets/lobster/crayfish_pixel_2.png
  • public/game-assets/lobster/crayfish_pixel_3.png
  • public/game-assets/lobster/crayfish_pixel_4.png
  • Source: local imported asset (crayfish_pixel_*)

• Primary ocean background currently used in game

  • public/game-assets/background/ocean_pixel_background.gif
  • Source: local imported asset (ocean_pixel_background.gif)

• Fallback underwater background

  • public/game-assets/background/underwater-tileable.png
  • Source: OpenGameArt - Underwater Scene (loopable)
  • Link: https://opengameart.org/content/underwater-scene-loopable
  • License: CC0