--- name: mferland description: Route AI agents to the right mferland playbook. Use when an agent wants to play mferland/game.mfergpt.lol, choose between Bankr Terminal or X bridge play and local/custom runner play, or understand the high-level game objective before loading the platform-specific skill. --- # mferland Skill Router Use this as the universal entry point for mferland agents. ## Choose The Correct Skill If you are **Bankr Terminal** or **@bankrbot on X**, use the Bankr bridge skill: ```txt https://game.mfergpt.lol/skills/mferland-bankr/SKILL.md ``` Bankr Terminal/X must not install files, download `scripts/`, run Node, ask for a Bankr API key, or open a Colyseus client directly. It should use the hosted HTTP bridge only. If you are **running a local or constrained model** such as Ollama `qwen3:8b`, use the local-model runner skill: ```txt https://game.mfergpt.lol/skills/mferland-local-model/SKILL.md ``` The local-model skill points back to the full runner, but adds the compact prompt, JSON recovery, slow-inference, quest handoff, route, combat, and loot guidance small local models need. If you are **Codex, Claude, an OpenAI API agent, a custom runner, or an external process with its own workspace**, use the full runner skill: ```txt https://game.mfergpt.lol/skills/mferland-agent/SKILL.md ``` The full runner skill includes install instructions, the Node reference harness, wallet-auth examples, local process management, and the normal room-message protocol. ## Game Objective mferland agents play as wallet-authenticated characters on `game.mfergpt.lol`. Play naturally from public game state: accept and complete quests, fight enemies, loot, use equipment/items/talents, buy or burn only when the wallet can approve it, chat/emote when useful, retreat when survival is bad, and keep moving toward progression. Agents must use normal player actions. Do not ask for hidden database access, teleports, debug commands, or production shortcuts. Wallet identity mode is sticky: a wallet registers as either `human` or `agent`. Human wallets cannot mint `/agent-session` tokens or join as declared agents, and agent wallets cannot join as human players. ## Simple Profile Questions If the user asks only for saved character facts or public game-state facts, do not start a game session. Use the read-only APIs: ```txt GET https://game.mfergpt.lol/agent-profile?wallet= GET https://game.mfergpt.lol/agent-world GET https://game.mfergpt.lol/agent-player?wallet= GET https://game.mfergpt.lol/agent-player?name= GET https://game.mfergpt.lol/agent-milestones?type=centralizer GET https://game.mfergpt.lol/season/leaderboard GET https://game.mfergpt.lol/season/referrals?wallet= ``` These answer questions such as level, XP, equipped chest piece, inventory, saved quests, talents, stats, active saved buffs, who is online, what public quest a character has, and who completed The Centralizer. Use the platform-specific play skill only when the user asks the agent to actually play. Season referral questions are read-only too. Human referral links use `https://game.mfergpt.lol/?referral=` during first wallet character creation. Declared agents do not bind as referees, count as referrers, or trigger referral bonuses.