360 Secure SkillHub
360 安全技能中心

Chess Buddy

该 Skill 面向国际象棋求助场景，支持棋盘图片、FEN、PGN 或自然语言描述。图片输入先按参考文档转成候选 FEN，并校验走子方、易位权、吃过路兵等要素；随后调用本地脚本运行 Stockfish 引擎分析，输出候选着法、关键变化线、局面计划与不确定性说明。不打包引擎，也不导入 python-chess 库。

tommyclawd
tommyclawd
热度 25.37
最近更新 2026-05-28 01:33:25
分类：知识依赖：代码执行应用场景：教学辅导 / 概念解释

安装技能

复制下面的话术发给你的 Agent，它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill： npx secure-skills-cli add chess-buddy
Metadata
name
chess-buddy
description
Stockfish-backed chess buddy for analyzing FEN positions and board images.
version
0.1.1
licenseAccepted
MIT-0
metadata
{"openclaw":{"requires":{"bins":["python3","stockfish"]},"homepage":"https://github.com/TommyClawd/chess-buddy-skill"}}

Chess Buddy

Use when the user wants chess help: board pictures, FEN validation, best moves, candidate lines, blunder review, or a training explanation.

License note

This ClawHub package is MIT-0 because ClawHub requires MIT-0 for all published skills. It is intentionally small: it treats Stockfish as an external executable dependency, does not bundle Stockfish, and does not import GPL Python libraries such as python-chess.

The richer GitHub source package is GPL-3.0-or-later because it imports python-chess and provides fuller FEN/PGN/SAN tooling: https://github.com/TommyClawd/chess-buddy-skill.

Guardrails

  • Do not assist cheating in active human games or rated online play. If the user is currently playing, offer post-game analysis or general principles only.
  • Image recognition is uncertain. Always verify side to move, board orientation, castling rights, en passant, and move number before treating an image-derived FEN as authoritative.
  • Image annotations are semantic, not self-explanatory. Do not assume an arrow/highlight is the played move; classify whether it is played move, engine recommendation, candidate line, last move, or mistake marker before explaining.
  • Engine output is evidence, not prose. Explain candidate moves in human terms and flag tactical vs strategic reasons.

Workflow

  1. Identify input type: image, FEN, PGN, or natural-language position.
  2. For images, follow references/image-to-fen.md; produce a candidate FEN, classify arrows/highlights, and verify the FEN before analysis.
  3. Run scripts/stockfish_fen.py engine-check once if the environment is new or engine availability is uncertain.
  4. Use scripts/stockfish_fen.py analyze --fen ... --multipv 3 --depth 12 for real engine analysis.
  5. Return concise analysis: position summary, top move(s), key line(s), plan, and uncertainty/verification notes.

Helper commands

python3 chess-buddy/scripts/stockfish_fen.py engine-check
python3 chess-buddy/scripts/stockfish_fen.py analyze --fen "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1" --multipv 3 --depth 12

If Stockfish is missing, report the missing dependency. Do not present non-engine commentary as chess analysis; call it notation validation, position description, or general coaching only.