Boardgame-Design-Hydra
# Boardgame Design Hydra A Claude Skill for designing tabletop games end to end: raw idea in, playable prototype pack out. It routes to genre-specific engines, runs balance and pacing checks, invents mechanics, writes rules, and plans playtests. It does not replace a designer. It exists so the boring, important questions get asked before.
Boardgame-Design-Hydra
Boardgame Design Hydra A Claude Skill for designing tabletop games end to end: raw idea in, playable prototype pack out. It routes to genre-specific engines, runs balance and pacing checks, invents mechanics, writes rules, and plans playtests. It does not replace a designer. It exists so the boring, important questions get asked before.
Author-TABARC-Code
---
Boardgame Design Hydra
A Claude Skill for designing tabletop games end to end: raw idea in, playable
prototype pack out. It routes to genre-specific engines, runs balance and
pacing checks, invents mechanics, writes rules, and plans playtests. It does
not replace a designer. It exists so the boring, important questions get
asked before the fun ones do.
---
Ok Note:
This is a great skill to do the manual boring crap of board game design, and pull the
small issues and little things we often miss and clog an initial Playtest. Its a
time save and helps those who are new to game design.
---
What it does
The skill classifies whatever the user brings it (a vague idea, a genre
question, a stuck prototype, a rules problem) and routes to the matching
part of the system rather than trying to do everything from one prompt.
- Ideas Go pipeline. The main workflow. Takes a raw game idea through
intake (max eight questions, with sensible defaults for anything skipped),
a design brief, four hard gates that must pass before any detail work
starts, a prototype pack, and a trade-off arbitration pass.
- Genre routing. Eighteen genre engines covering 18xx, 4X, puzzle,
abstract, Ameritrash, party, deck-building, worker placement, area
control, co-op, hidden role, legacy, roll-and-write, wargame, dexterity,
trivia, racing, and general strategy, plus a dedicated eurogame theory
file covering mechanics taxonomy, the German market and awards culture,
and community dynamics.
- Utility engines. Balance, pacing, player psychology, randomness and
variance, thematic integration, UX and component design, production,
pitching, critical analysis of existing games, and a design-reduction
pass for cutting a prototype back to what actually works.
- Mechanic invention. A constraint-mutation framework for generating
genuinely new mechanics rather than reskinned familiar ones.
- Referee rules. A fixed lens-priority hierarchy (intent, then
friction, then balance, then core loop, then component economy, then
randomness, then interaction, then market fit) that arbitrates
conflicting design pressures the same way every time, plus kill criteria
for recommending a project be shelved after three unproductive
iterations.
- Playtest system. Fourteen fictional playtest personas, written as
drop-in NPC character sheets with voice, schedule constraints, and
table quirks, for simulating how a session actually plays before a real
group sees it.
- Control plane. A beginner wizard for turning a vague idea into a
structured brief, an autopilot runner for taking a brief straight to a
prototype pack, and a dashboard view with a risk register.
Status
Implemented and in active use as a Claude Skill: the router, all eighteen
reference files, ten output templates, and the JSON schema for the design
brief are complete and internally consistent. It has not been run through
a formal trigger-accuracy evaluation or a benchmarked test suite; feedback
on it so far is qualitative, from use rather than from a scored eval loop.
How it works
SKILL.md is the router. It classifies the incoming request against a
table, then either runs the Ideas Go pipeline directly or points to the
relevant reference file for deeper genre or engine knowledge. Reference
files are only read when the routing table says they're needed, which
keeps the skill's footprint small until a specific domain is actually in
play. Output goes through one of ten templates, so a prototype pack, a
rules draft, and a playtest report all come out in a fixed, comparable
shape rather than freeform prose each time.
``
boardgame-design-hydra/
SKILL.md router + Ideas Go pipeline + referee rules
references/
genres.md 18 genre-specific design engines
eurogames.md deep eurogame theory
engines.md balance, pacing, psychology, randomness, etc.
mechanic-invention.md constraint-mutation framework
referee.md lens priority + kill criteria
playtest-system.md playtest methodology
personas.md 14 playtest personas
pipeline.md wizard, autopilot runner, control plane
writing-doctrine.md house prose style
templates/ 10 output-format templates
contracts/
design_brief_v1.schema.json
Requirements
Runs as a Claude Skill inside Claude.ai, Claude Code, or Cowork. No
external runtime, dependency, or account is required; it is markdown and
one JSON schema, read by Claude at trigger time.
Installation
Place the boardgame-design-hydra/ folder wherever the host surface loads
skills from (for Claude.ai and Claude Code, typically alongside other
installed skills). No build step.
Usage
Bring it a raw idea ("I want a fifteen-minute party game about lying to
your relatives") and it runs the Ideas Go pipeline: intake questions where
needed, a design brief, the four hard gates, then a prototype pack with
core loop, win condition, component list, rules v0.1, and a playtest plan.
Bring it a narrower question instead ("why does my worker-placement game
have a runaway leader problem", "invent a new mechanic for hidden
information", "write a rules section for combat resolution") and it routes
straight to the matching engine or reference file rather than running the
full pipeline.
Configuration
Nothing to configure. The skill's only external contract isdesign_brief_v1.schema.json`, which defines the shape of a design brief
(title, elevator pitch, player count, duration, interaction mode, core
verbs, randomness level, success metrics) and is used internally to keep
briefs consistent across a session.
Known limitations
- The referee's lens hierarchy and kill criteria are fixed by design and
don't currently take a user override; if a project's priorities
genuinely don't match the built-in order (learnability > experience >
fairness > production > novelty), that tension has to be worked around
in conversation rather than reconfigured.
- The fourteen playtest personas are a fixed, hand-written set. They give
useful variety in feedback style but are not a substitute for an actual
playtest group, and the skill doesn't claim otherwise. - Other play tester skillls are on way
- No automated eval set or benchmark exists yet for trigger accuracy or
output quality; testing so far has been manual.
Author
TABARC-Code
Bring it a raw idea ("I want a fifteen-minute party game about lying to
your relatives") and it runs the Ideas Go pipeline: intake questions where
needed, a design brief, the four hard gates, then a prototype pack with
core loop, win condition, component list, rules v0.1, and a playtest plan.
Bring it a narrower question instead ("why does my worker-placement game
have a runaway leader problem", "invent a new mechanic for hidden
information", "write a rules section for combat resolution") and it routes
straight to the matching engine or reference file rather than running the
full pipeline.
Runs as a Claude Skill inside Claude.ai, Claude Code, or Cowork. No
external runtime, dependency, or account is required; it is markdown and
one JSON schema, read by Claude at trigger time.