ワンクリックでManusで任意のスキルを実行

optimize-pyray-game

スター0

フォーク0

更新日2026年7月20日 04:48

Profile and optimize a verified pyray/raylib game using representative scenes and explicit budgets. Use only for measured performance work after correctness gates pass.

インストール

Codex または Claude でインストールこの Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。

Manusで実行

ソース

DrHepa

DrHepa/rpg-world-forge

GitHub リポジトリを開く Creator のリポジトリを見る

ダウンロード

Manusで実行

Optimize a pyray game

Scope

Optimize one measured G31 bottleneck in its game-owned implementation surface. Do not combine unrelated optimizations or change contracts/features.

Inputs

Absolute FORGE_ROOT and external GAME_ROOT.
Exact G30 source-tree/runtime/platform/catalog/bundle identities and baseline, with correctness gates passing and one budget miss or documented goal.
One reproducible scene, measured metric, numeric acceptance threshold, target platform/backend, and owning G10-G25 writable surface.

Outputs

At most one game-owned implementation optimization and focused regression tests.
Forge-owned before/after distributions, threshold result, optimized-tree identity, invalidated prior G30 evidence, and explicit handoff to a separate G30 invocation.

Invariants

Keep src/isoworld, locks, bundles, public contracts, determinism, visual semantics, path validation, and resource teardown intact.
Stop and re-enter the owning phase if a contract/API change is required.
Never copy AGENTS.md, .agents, Forge skills, or phase evidence into GAME_ROOT.

Workflow

Read the performance gate in docs/GAME_IMPLEMENTATION_PHASES.md.
Record a reproducible baseline scene, platform, backend, and bundle hash.
Measure simulation, render planning, drawing, load time, memory, and CFFI calls.
Optimize in order: culling, atlases, precomputed commands, caching, batching.
Apply only an implementation-level change inside the measured owning surface.
Compare before/after medians and tails, run focused/shared regressions and the clean-game audit, then hand the exact optimized identity to a separate G30.

Completion

Complete only when the predeclared metric reaches its numeric threshold without regression and focused/shared tests plus the clean-game audit pass. Invalidate old G30 evidence and require a separate G30 invocation before G32; never claim or write G30 evidence from this skill.

このリポジトリの他の Skills

同じリポジトリ

animate-3d-asset-with-blender

DrHepa/rpg-world-forge

Animate one reviewed rig through bounded Blender MCP operations under GPT orchestration. Use for specified clips and loop evidence only, not rig changes, final GLB export, refinement, or QA approval.

2026-07-200

define-asset-bibles

DrHepa/rpg-world-forge

Define and approve target-scoped visual and audio bibles from a locked worldpack. Use for M5 art direction only, not inventory derivation, asset generation, processing, or QA.

2026-07-200

derive-asset-inventory

DrHepa/rpg-world-forge

Derive a complete target-scoped M5 asset inventory from locked canon and approved bibles. Use for requirements and semantic slots only, not specifications, generation, or processing.

2026-07-200

design-3d-asset-references

DrHepa/rpg-world-forge

Produce a coherent reference-image set for one approved 3D asset specification through OpenAI Image. Use for 3D design evidence only, not modeling, rigging, animation, GLB export, or QA.

2026-07-200

forge-world-release

DrHepa/rpg-world-forge

Compile, package, and verify one immutable runtime bundle from an approved world-authoring repository. Use only for the world-release phase; game import is a separate skill.

2026-07-200

model-3d-asset-with-blender

DrHepa/rpg-world-forge

Model one specified 3D asset through bounded Blender MCP operations under GPT orchestration. Use for geometry, UVs, and material slots only, not rigging, animation, final export, or QA approval.

2026-07-200

Optimize a pyray game

Scope

Optimize one measured G31 bottleneck in its game-owned implementation surface. Do not combine unrelated optimizations or change contracts/features.

Inputs

Absolute FORGE_ROOT and external GAME_ROOT.

Exact G30 source-tree/runtime/platform/catalog/bundle identities and baseline, with correctness gates passing and one budget miss or documented goal.

One reproducible scene, measured metric, numeric acceptance threshold, target platform/backend, and owning G10-G25 writable surface.

Outputs

At most one game-owned implementation optimization and focused regression tests.

Forge-owned before/after distributions, threshold result, optimized-tree identity, invalidated prior G30 evidence, and explicit handoff to a separate G30 invocation.

Invariants

Keep src/isoworld, locks, bundles, public contracts, determinism, visual semantics, path validation, and resource teardown intact.

Stop and re-enter the owning phase if a contract/API change is required.

Never copy AGENTS.md, .agents, Forge skills, or phase evidence into GAME_ROOT.

Workflow

Read the performance gate in docs/GAME_IMPLEMENTATION_PHASES.md.

Record a reproducible baseline scene, platform, backend, and bundle hash.

Measure simulation, render planning, drawing, load time, memory, and CFFI calls.

Optimize in order: culling, atlases, precomputed commands, caching, batching.

Apply only an implementation-level change inside the measured owning surface.

Compare before/after medians and tails, run focused/shared regressions and the clean-game audit, then hand the exact optimized identity to a separate G30.

Completion

name	optimize-pyray-game
description	Profile and optimize a verified pyray/raylib game using representative scenes and explicit budgets. Use only for measured performance work after correctness gates pass.