game-ideation

Game Ideation AI Agent Skills

A structured, AI-agent-supported workflow that takes a raw game idea through to a complete, implementation-grade Game Concept — guided by the methodology of Daniel Dumont's six-part workshop "Von der Idee zum Konzept" (Making Games Magazin, 2010–2011).

What this skill set does

Game design documentation tends to fail in one of two ways: either it stays a vague pitch that nobody can build from, or it grows into a sprawling document that nobody can read. Dumont's method solves both by walking the author through three stages of increasing detail, with explicit gates between stages and discipline rules that keep each stage focused.

This repository turns that method into a set of AI agent skills. Each skill:

• Acts as an experienced co-designer — not a passive scribe, but an active partner that questions weak points, surfaces design risks, and contributes creative impulses
• Knows its stage's rules (length limits, required components, anti-patterns) from the Dumont workshop summaries
• Maintains continuity across stages — earlier decisions are respected and inherited by later stages
• Captures out-of-stage details in side-notes so nothing is lost when the author drills into something prematurely

The result: a coherent set of design documents that grow from a half-page vision into a full implementation-ready concept, without the author losing the overview at any point.

A word on the methodology

Dumont's workshop is from 2010/2011 — methodologically rooted in the era of classic waterfall planning that German studios like Ascaron (Dumont's home) practiced at the time. We use it deliberately, with eyes open.

What is timeless:

Element	Why it stays valid
Vision → Exposé → Concept staging	Mirrors any structured design work
Precision discipline	Every gap in the spec costs time later
Change management as a practice	Untracked decisions always come back to bite
Feature chapters covering edge cases, dependencies, balancing values	The questions you skip are the bugs you ship

What feels 2010s-era:

Element	Why it's dated
Strongly waterfall: "design completely first, then build"	Modern practice ships earlier and iterates
Implicit assumption of a full team (producer, QA, dedicated concept author)	Assumes infrastructure most solo devs lack
Paper-design over prototyping	Today's vertical-slice-first approach is more common
No mention of living docs, wikis, modern tooling	Modern teams use Notion, Confluence, and similar

Why this fits AI-assisted solo development:

Modern indie and AAA practice is heavily prototype-driven — vertical slice first, write the spec afterward. That works when you have a team carrying verbal decisions around. Solo developers working with AI assistance are in a different situation: there is no team to absorb undocumented decisions, and the AI agent's output quality is bounded by the precision of its input.

Dumont's strictness — write it down, write it precisely, gate each stage — turns out to be a much better match for this case than modern-agile practice. Documentation is the solo developer's memory. It is also the AI agent's input contract.

How it works — the three-stage flow

Dumont's core insight is three stages of conception, each gated by review and viability: