メニュー
Use SkullScope physics diagnostics with focused queries, mandatory query logging, and GPT-read data-size accounting.
デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。
インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。
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。
直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。
npx skills add https://github.com/skullbonez/SkullbonezCore --skill skore-skullscopeコマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。
ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。
Run SkullbonezCore's persistent MASTER-PLAN implementation queue on a deterministic Night Runner branch, either serially in the main agent or as the integration owner for the parallel orchestrator. Use when the user asks for the orchestrator, night runner, nightrunner, overnight runner, queued plan runner, MASTER-PLAN runner, or asks Codex/Antigravity to complete one or more Agentic/Plans items, then continue through blockers, validate, commit, and push each accepted plan.
Run a hard-nosed Carmack-inspired engine quality evaluation. Use when the user asks whether SkullbonezCore or another game engine is robust, performant, encapsulated, deterministic, commercially credible, suitable as a physics engine, or whether an elite systems programmer would rate, buy, or use it.
Audit SkullbonezCore source comments for accurate ownership, invariants, hazards, lifetimes, units, and caller contracts.
SOC 職業分類に基づく
SKILL.md を表示中
| name | skore-skullscope |
| description | Use SkullScope physics diagnostics with focused queries, mandatory query logging, and GPT-read data-size accounting. |
Use this skill whenever a physics investigation needs runtime diagnostics, rest-state analysis, energy accounting, contact/island inspection, or a comparison between a raw physics log and queryable SkullScope output.
Do not ingest whole physics CSV, NDJSON, or SQLite files unless the user explicitly asks for raw logs. Generate or reuse a deterministic --physics-diag trace, query it with tools\physics_query.bat, and expose only bounded query results to the model.
Every handoff or final answer after using SkullScope must print:
tools\physics_query.bat query run,Raw artifact size and GPT-read size are different numbers. The GPT-read size is only the query output text shown to the model, not the NDJSON trace or SQLite cache sitting on disk.
tools\validate_build.bat Debug
Debug\SKULLBONEZ_CORE.exe --renderer dx12 --scene SkullbonezData\scenes\scene.scene.json --physics-diag Debug\scene.physicsdiag.ndjson --vsync off
Get-Item Debug\scene.physicsdiag.ndjson
Get-Item Debug\scene.physicsdiag.sqlite
tools\physics_query.bat Debug\scene.physicsdiag.ndjson summary
tools\physics_query.bat Debug\scene.physicsdiag.ndjson events --limit 20
tools\physics_query.bat Debug\scene.physicsdiag.ndjson contacts --frame <frame> --body <body>
tools\physics_query.bat Debug\scene.physicsdiag.ndjson body <body> --frames <start>:<end> --limit 50
tools\physics_query.bat Debug\scene.physicsdiag.ndjson island <id> --frame <frame>
tools\physics_query.bat Debug\scene.physicsdiag.ndjson summary | Tee-Object Debug\scope_summary.json
(Get-Item Debug\scope_summary.json).Length
(Get-Content -Raw Debug\scope_summary.json).Length
--limit, --frame, --frames, --body, --type, or --severity.Summarize the diagnosis in plain language, then include a compact audit block:
SkullScope artifacts:
- Trace: Debug\scene.physicsdiag.ndjson, 22,191,227 bytes
- SQLite cache: Debug\scene.physicsdiag.sqlite, 4,923,392 bytes
Queries read by GPT:
- tools\physics_query.bat Debug\scene.physicsdiag.ndjson summary
Output read: 5,842 chars, 5,842 bytes
- tools\physics_query.bat Debug\scene.physicsdiag.ndjson body marble --frames 320:340 --limit 50
Output read: 12,104 chars, 12,104 bytes
Total GPT-read SkullScope output: 17,946 chars. Raw trace was not read by GPT.
Documentation-only changes require no validation.
For SkullScope tool or physics diagnostic changes, run this as the targeted pre-commit/PR validation gate:
tools\validate_physics.bat