skills AGENTS.md

Repo Guide

This repository is a small skill and template library with one explicit domain package: gamedev/.

Read Order

1. Read this file for the repo map.
2. Read docs/context-management.md if the task will span multiple edits or documents.
3. Read docs/gamedev-workflow.md for the canonical full-run vs step-by-step routing when the task is game-related.
4. Read docs/gamedev-specialist-handoffs.md when the task needs runtime-, engine-, UI-, asset-, or QA-specific guidance.
5. Load only the layer you need:
   • core/ for shared, domain-agnostic workflows
   • gamedev/ for game-specific workflows, standards, and templates
   • templates/ for shared document skeletons
   • standards/ for shared conventions
   • productivity/ for optional helper utilities

Boundaries

• core/ contains only workflows that still make sense if gamedev/ disappears.
• docs/ contains repository guidance and explanatory reference docs, not templates or standards.
• templates/ contains reusable shared skeletons only.
• standards/ contains compact shared conventions only.
• gamedev/ contains game-specific skills, standards, and templates.
• gamedev/_archive/ contains good but non-core gamedev material kept outside the active workflow path.
• productivity/ contains optional utilities and must not be a dependency of core/.

Defaults

• Prefer moving useful game-specific material into gamedev/ over forcing it into the shared core.
• Keep a single canonical template per shared artifact type.
• Prefer short, direct workflows over ceremony-heavy routing skills.
• Keep gamedev/ platform-agnostic; do not copy browser-runtime doctrine from specialist plugins into the generic flow.

Gamedev Flow

• The active gamedev/ path now has two core lanes plus one optional demo extension. Preproduction: setup-engine -> map-systems -> design-system -> prototype. Production bridge: bootstrap-project -> implement-system -> assemble-mvp -> playtest-and-tune. Demo extension: prepare-demo -> design-system/implement-system -> assemble-mvp -> playtest-and-tune.
• Typical small-game handoff: lock the stack, map systems, design the first MVP systems, prototype only where risk is real, then bootstrap the scaffold and move into implementation and playable-loop assembly. When the milestone changes from prove the loop to show a credible demo, run prepare-demo before broad UI, asset, or presentation work.
• The flow is platform-agnostic. Browser projects may overlay Game Studio for runtime, UI, and asset depth, and Browser Use or another browser specialist for local browser inspection and playtest evidence. Non-browser projects should use engine-native specialists or official engine docs without changing the core flow ownership.
• bootstrap-project may happen as soon as docs/technical-preferences.md is stable and the repository needs a runnable codebase, but do not skip map-systems or design-system when system scope is still unclear.
• implement-system is for one approved system at a time and should push the systems index to implemented when code exists.
• assemble-mvp is only for wiring multiple implemented systems into one coherent playable loop and should push verified systems to integrated.
• prepare-demo is for defining the first audience-facing demo contract from a proven slice and should turn vague polish wishes into explicit presentation, asset, and QA handoffs.
• playtest-and-tune is for focused tuning after a playable slice exists.
• Keep prototype code isolated under prototypes/; do not quietly promote it into production.
• Prefer a small playable loop over broad feature coverage.