SuperAPI AI Directory

game-concept

General & other

View repository

Updated 10-16 09:30

About this skill

Game Concept Design

Decide what kind of game to make, without writing code, numeric tables, or scene-by-scene level scripts.

Read concept-method.md. The input must include SOURCE_BIBLE.md and PRODUCT_BRIEF.md; if either is missing, stop and state the gap, and do not fill it in on behalf of upstream.

The artifact language is locked by PRODUCT_BRIEF.md ; when not locked, follow the conversation language, and do not default to producing Chinese.

Decision boundaries

Directly inherit the platform, input, audience, rating, duration, target market, and non-goals already locked by the brief; when they are found to clearly conflict with the original work, return to the main entry to revise, and do not silently change values during the concept stage. Handle the original work's language, cultural context, target market, and interface language separately, and do not automatically infer one from another.

Interactive narrative is a mature gameplay form, and its standards are not lowered just because it is carried by reading and dialogue; its verbs can be reading the situation, pressing with questions, comparing information, taking a stance, concealing, handing over, and bearing the echoes.

The concept must also determine the validation form: when it can be expressed with mature interaction grammar, state "existing grammar + how the original work changes it"; when the core depends on real-time feel, space, line of sight, physics, or unique operations, clearly go with a custom greybox. Do not treat text prototyping as the unified prerequisite layer for all experiences, and do not let templates decide the gameplay for the original work just because templates are available.

Comparison scope

First list the brief's locked dimensions and still-open dimensions, and compare only directions that can reveal real trade-offs; do not pad the number of concepts or difference dimensions. Only examine credible alternatives when the direction is still open; when the user has already chosen a direction or no valid alternative exists, explain why, switch to validating that direction's risks, and do not reopen the choice.

Selection

First eliminate one by one according to concept-method's hard vetoes, without calculating a total score, then choose according to its comparison dimensions.quick automatically select the direction with the strongest evidence;director only give a recommendation for directions not yet decided and wait for the user to choose.

Output

Generate one concepts/CONCEPT.md, containing only:

  1. brief locked values and open dimensions;
  2. the experience promise and its observable phenomena and failure phenomena;
  3. compact precedent explanation;
  4. candidate concept card: main type,experienceProfile, borrowed/rejected gameplay precedents, core actions and loop, pressure, failure, world response; explain how the player identity forms repeated actions, execution resistance, and observable change, without requiring independent fixed key names. signature_command Write the value according to the method "identity-specific command" or N/A; finally explain how the original work's tension is enacted by the player firsthand, the signature image, the minimum validation slice, the maximum risk, and the veto observation;
  5. hard veto results, key trade-offs, recommendation, and selection status;
  6. the selected direction's player agency contract and non-negotiable experience promise;
  7. minimum validation questions and open questions.

narrative-led or hybrid The narrative layer must also choose a branching structure grammar and explain the content budget. Before handoff, confirm that the selected direction can be made into a complete slice within the brief's scope.