ワンクリックでManusで任意のスキルを実行

gundam-bot-bench

スター29

フォーク10

更新日2026年7月15日 10:00

Run reproducible Gundam bot self-play, diagnose strategy or engine failures, and compare heuristic candidates with tools/bot-bench. Use for bot fuzzing, strategy changes, deadlocks, illegal moves, or win-rate evaluation.

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

TheCardGoat

TheCardGoat/tcg-engines

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

Gundam Bot Bench

The bench is a behavioral sensor, not a substitute for focused tests. Read tools/bot-bench/README.md, the CLI help in scripts/bench.ts, and the current strategy registry before choosing arguments.

Reproducible Run

From submodules/gundam/tools/bot-bench:

vp run bench -- \
  --p1 greedy-legal --p2 value-ranked \
  --p1-deck ef-starter --p2-deck seed-aggro \
  --matches 100 --seed-base candidate-1 \
  --out reports/baseline.json

Use the same strategies, decks, match count, and seed base for candidate comparisons. Then run:

vp run diff -- \
  --baseline reports/baseline.json \
  --candidate reports/candidate.json \
  --for p1 --out reports/diff.json

Workflow

Run focused engine automation and bot-bench tests first.
Capture a baseline report with explicit parameters.
Form one hypothesis from termination reasons, illegal-move errors, family attempt/success rates, or repeatable strategy behavior.
Make one typed strategy or engine change.
Add a focused regression test for every illegal move, deadlock, or engine exception found by self-play.
Re-run both seats with the same seed base and compare reports.
Keep a heuristic only when the relevant behavior improves without masking a family regression. A small win-rate swing alone is insufficient.

Boundaries

Read current CLI flags; do not use obsolete --games, --decks, or --seed aliases.
Full reports are reproducible artifacts and normally stay uncommitted.
UI exposure belongs in ../agnostic-simulator/apps/multi-game-simulator/src/games/gundam, not a simulator app inside this workspace.
Stop after one hypothesis unless the user explicitly requests an autonomous loop with acceptance and stop criteria.

Report parameters, seed base, baseline/candidate results, family deltas, focused tests, and a keep/reject conclusion.

このリポジトリの他の Skills

同じリポジトリ

lorcana-cards

TheCardGoat/tcg-engines

Single owner for everything related to a specific Disney Lorcana card in `packages/lorcana/lorcana-cards` — authoring, debugging, triaging, generating, and any bounded engine or types extension required to make the printed text work. Use whenever the unit of work is "this card".

2026-07-1829

lorcana-rules

TheCardGoat/tcg-engines

Explain and cite Disney Lorcana rules using indexed references, and produce implementation-ready behavior constraints for card development. Use when answering mechanic questions, resolving card interactions, or deriving test/implementation constraints from official rules.

2026-07-1829

adapt

TheCardGoat/tcg-engines

Adapt designs to work across different screen sizes, devices, contexts, or platforms. Implements breakpoints, fluid layouts, and touch targets. Use when the user mentions responsive design, mobile layouts, breakpoints, viewport adaptation, or cross-device compatibility.

2026-07-1529

animate

TheCardGoat/tcg-engines

Review a feature and enhance it with purposeful animations, micro-interactions, and motion effects that improve usability and delight. Use when the user mentions adding animation, transitions, micro-interactions, motion design, hover effects, or making the UI feel more alive.

2026-07-1529

arrange

TheCardGoat/tcg-engines

Improve layout, spacing, and visual rhythm. Fixes monotonous grids, inconsistent spacing, and weak visual hierarchy. Use when the user mentions layout feeling off, spacing issues, visual hierarchy, crowded UI, alignment problems, or wanting better composition.

2026-07-1529

audit

TheCardGoat/tcg-engines

Run technical quality checks across accessibility, performance, theming, responsive design, and anti-patterns. Generates a scored report with P0-P3 severity ratings and actionable plan. Use when the user wants an accessibility check, performance audit, or technical quality review.

2026-07-1529

Gundam Bot Bench

Reproducible Run

From submodules/gundam/tools/bot-bench:

vp run bench -- \ --p1 greedy-legal --p2 value-ranked \ --p1-deck ef-starter --p2-deck seed-aggro \ --matches 100 --seed-base candidate-1 \ --out reports/baseline.json

Use the same strategies, decks, match count, and seed base for candidate comparisons. Then run:

vp run diff -- \ --baseline reports/baseline.json \ --candidate reports/candidate.json \ --for p1 --out reports/diff.json

Workflow

Run focused engine automation and bot-bench tests first.

Capture a baseline report with explicit parameters.

Form one hypothesis from termination reasons, illegal-move errors, family attempt/success rates, or repeatable strategy behavior.

Make one typed strategy or engine change.

Add a focused regression test for every illegal move, deadlock, or engine exception found by self-play.

Re-run both seats with the same seed base and compare reports.

Keep a heuristic only when the relevant behavior improves without masking a family regression. A small win-rate swing alone is insufficient.

Boundaries

Read current CLI flags; do not use obsolete --games, --decks, or --seed aliases.

Full reports are reproducible artifacts and normally stay uncommitted.

UI exposure belongs in ../agnostic-simulator/apps/multi-game-simulator/src/games/gundam, not a simulator app inside this workspace.

Stop after one hypothesis unless the user explicitly requests an autonomous loop with acceptance and stop criteria.

Report parameters, seed base, baseline/candidate results, family deltas, focused tests, and a keep/reject conclusion.

name	gundam-bot-bench
description	Run reproducible Gundam bot self-play, diagnose strategy or engine failures, and compare heuristic candidates with tools/bot-bench. Use for bot fuzzing, strategy changes, deadlocks, illegal moves, or win-rate evaluation.