Gamemode Designer

Gamemode Designer (the foundry)

Take an idea all the way to a fun, re-engaging server. One command, /mcwrench:create, runs three internal phases. Plan and write locally only. Never run live server commands, RCON, or panel writes — that is a future operator mode, out of scope here. Never invent plugin config keys; map every mechanic to a real plugin or flag it as custom, and fetch exact keys with learn-plugin-docs.

How fun works (bake this in)

A gamemode is fun when it has a clear Challenge → Action → Reward core loop, a nested or escalating loop for retention (prestige, seasons, a ladder to climb), and a twist on a familiar base rather than a clone. Originality comes from combining existing archetypes + exactly one bending twist + a theme — evolve, do not copy.

Phase 1 — IMAGINE (only when the input is vague)

If the user gives nothing, a vibe, or "something fun", generate 3 to 5 original concepts. For each, fill: name, one-line pitch, base archetype(s) fused, the one twist, core loop, progression, conflict + stakes, target audience, fun score, feasibility verdict + the plugin stack, and a novelty note. Build concepts from references/mechanic-taxonomy.md (the 9 axes + twist bank). Score with references/fun-feasibility-rubric.md. Propose the top 1 to 3 and let the user pick, or crown the highest fun among the buildable-now options.

If the user already has a concrete idea ("a server where you age and die of old age", or a concept from a previous IMAGINE), skip to DESIGN.

Phase 2 — DESIGN the chosen concept

Lock it down: the core loop in one sentence, the single twist, the full plugin stack (each mechanic mapped to a real plugin), and the glue the twist needs (a config setting, a Skript, or a custom plugin). Sanity-check feasibility: stack exists, performance risk, abuse surface (alt-farming, grief, combat-logging, dupes), EULA pressure. Give a verdict: Config-only, Skript-glue, or Needs a custom plugin (park the last for later). Prefer the highest-fun idea that is Config-only or Skript-glue — that is one that "could actually work".

Phase 3 — CREATE the local build plan / starter pack

Route to the existing skills and assemble one plan:

• Stack + docs → gamemode-stacks for the closest base archetype, then pre-pull every plugin's docs with learn --stack <closest> / learn-plugin-docs.
• Server setup → new-server-bootstrap (software, Java, startup, EULA, starter configs).
• Ranks → a themed ladder handed to permissions-helper for LuckPerms tracks.
• Brand → server-branding for the MOTD, rank prefixes, and store/Discord copy in the right format per plugin.
• The twist → a concrete spec (which plugin/Skript, what it does, the config or event hooks).
• Audit → an audit-config pass over the proposed config.

Emit it as a single readable starter pack: concept, plugin list, the twist spec, ranks, MOTD, and a build order. Write nothing to a live server. Local starter files only if the user explicitly confirms (and only once config generators exist).

Guardrails

• Local only. No RCON, no panel API writes, no console commands, no live testing. Those are v2.
• Never invent config keys, permission nodes, or plugin behaviour. Fetch with learn-plugin-docs.
• Keep the economy/store EULA-safe; flag pay-to-win pressure a twist might create.
• One twist per concept. Two twists usually means two muddy gamemodes.

References

• references/mechanic-taxonomy.md — the 9 combination axes + the twist bank.
• references/fun-feasibility-rubric.md — the fun rubric, the feasibility check, the verdict ladder.
• references/example-concepts.md — worked hybrid examples (Tidewell, Faultline, Ageless) + seeds.