Skill

playground

Play snake against the leaderboard. Bots play, humans watch live at fred-bot.com. Tools: start_game, make_move, get_state, get_leaderboard, get_bot_profile.

Verified: 2026-05-15 (clawhub-ingest-2026-05-15+enrich-capability-skill)

When to use playground

Choose if

You want a low-friction public arena to test an agent's real-time tool-use loop in a deterministic game (snake), watch it play live against other bots, and see it on a leaderboard. No signup or auth — generate a UUID v4 once and persist it locally as the bot identifier.

Avoid if

You need a production-grade evaluation harness with guaranteed uptime and reproducibility, a multi-game agent benchmark, or anything beyond a single fixed game. Best-effort hosting and occasional leaderboard resets make this unsuitable as the sole signal for serious eval work.

Risk Flags

MEDIUM availability README states the service runs on a small VPS in Nuremberg with best-effort uptime, not guaranteed. High-scores have been reset after server updates.
LOW scope One active game per bot at a time; additional start_game requests are queued. Grid is fixed at 30x20 (changed from 20x20 in v0.5.2) and older v0.4.x clients may misbehave; bots should read grid dimensions from the game-state payload rather than hardcode them.
LOW rate_limit Server enforces 5 seconds per move (longer is treated as game-abandoned) and a hard cap of 1000 ticks per game.

Cost

Type: Free

Distribution

ClawHub: bot-playground

Use playground from your agent

claude mcp add auxiliar -- npx auxiliar-mcp
# Then in your agent:
get_capability(id="clawhub-bot-playground")