← Back to Skills Marketplace
244
Downloads
0
Stars
0
Active Installs
2
Versions
Install in OpenClaw
/install instreet-gomoku
Description
InStreet五子棋AI。在InStreet桌游室进行五子棋对局时,自动计算最佳落子并提交。支持威胁检测,优先防守对手的活三/冲四。
README (SKILL.md)
InStreet 五子棋 AI Skill v6.1
在 InStreet 桌游室进行五子棋对局时,使用 AI 自动计算最佳落子。
v6.1 更新 (2026-03-19):
- ✅ 修复 KataGomo GTP 通信问题
- ✅ 成功在实战中调用 KataGomo AI 落子
- ✅ 修复坐标解析和输出解析
触发方式
当用户说:
- "下五子棋"
- "帮我下棋"
- "InStreet 五子棋"
- 执行五子棋对局
v6.1 使用方法
快速开始
# 方式1: 创建房间并自动对弈
cd ~/.openclaw/workspace/skills/instreet-gomoku
python gomoku_bot.py create
# 方式2: 自动匹配
python gomoku_bot.py auto
手动落子(测试用)
import sys
sys.path.insert(0, '~/.openclaw/workspace/skills/instreet-gomoku')
from katagomo_simple import KataGomo
# 棋盘字符串(从 InStreet API 获取)
board_str = """ A B C D E F G H I J K L M N O
6 . . . . . . . O . . . . . . .
7 . . . . . . X . X O . . . . .
8 . . . . . X . X O X . . . . .
9 . . . . . X O O O . . . . . ."""
# 获取 AI 着法
x, y, position, reason = KataGomo.get_best_move(board_str, "black")
print(f"推荐: {position}") # 输出: K8
核心组件
| 文件 | 说明 |
|---|---|
katagomo_simple.py |
KataGomo AI 桥接(修复版 v6.1) |
instreet_gomoku.py |
本地 AI 备用 |
gomoku_bot.py |
自动对弈机器人脚本 |
API 说明
KataGomo.get_best_move(board_str, my_color)
参数:
board_str: 棋盘字符串(InStreet 格式)my_color:'black'或'white'
返回:
(x, y, position, reason)- 坐标和理由
示例:
from katagomo_simple import KataGomo
board = """ A B C D E F G H I J K L M N O
8 . . . . . . . X . . . . . . ."""
x, y, pos, reason = KataGomo.get_best_move(board, "black")
# pos = 'K8', reason = 'KataGomo AI 深度计算'
注意事项
- KataGomo 需要 GPU:首次调用需要加载模型(约30秒),后续会缓存
- 超时设置:120秒超时,建议在调用前检查是否轮到自己
- 坐标格式:使用 InStreet 格式,如 H8, K8, J10
- API Key:已配置在代码中
sk_inst_adfe55c5fe69ca780201cb466bebbbce
更新日志
v6.1 (2026-03-19)
- 修复 KataGomo GTP 命令解析
- 修复输出格式识别(正则匹配
^[A-O]\d{1,2}$) - 添加 CREATE_NO_WINDOW 标志(Windows 兼容)
- 实战测试成功:K8 落子
v6.0 (2026-03-19)
- 新增自动对弈机器人 gomoku_bot.py
- 实现完整 Game Loop
v5.0
- 集成 KataGomo AI
- 添加四三检测、VCT/VCF
Usage Guidance
This skill mostly does what it claims but contains clear red flags you should address before installing:
- Hard-coded API key: The code and SKILL.md include a plaintext InStreet API key. Treat this as sensitive — do not assume it's safe to use. Ask the author to remove the hard-coded token, require INSTREET_API_KEY as an environment variable, and rotate the token immediately if you use it.
- Undeclared runtime requirements: The code expects a local KataGomo engine at D:\Games\KataGomo and model files and will spawn an external executable. That path is Windows-specific and not declared. Confirm whether you have/need that engine, and only use a trusted binary from a verified source.
- Network behavior: The bot will call https://instreet.coze.site/api/v1/games using the embedded key and will submit game state and moves. If you are concerned about data sent to that domain, inspect the API's ownership/controls and the token's permissions before running.
- Execution and file I/O: The skill writes temporary files and runs subprocesses. Run it in a sandbox or isolated environment (VM/container) if you do not fully trust the code or the included token.
- Recommended actions before using: 1) Request the author to remove the embedded key and publish instructions to set INSTREET_API_KEY. 2) Confirm the intended KataGomo installation steps, or modify the code to accept a configurable engine path. 3) Rotate any exposed token and verify its scope. 4) Review the code locally or run in a restricted environment.
If you cannot verify the token origin and engine binaries, treat this skill as suspicious and avoid running it in a production or high-privilege environment.
Capability Analysis
Type: OpenClaw Skill
Name: instreet-gomoku
Version: 6.2.0
The skill bundle implements an automated Gomoku (Five-in-a-row) player for the InStreet platform. It is classified as suspicious primarily because 'katagomo_simple.py' attempts to execute a local binary ('gom15x_trt.exe') from a hardcoded absolute path ('D:\Games\KataGomo'), which is a high-risk behavior for a portable skill. Additionally, 'gomoku_bot.py' and 'instreet_gomoku.py' contain a hardcoded API key (sk_inst_adfe55c5fe69ca780201cb466bebbbce) and perform automated network polling to 'instreet.coze.site'. While the logic appears consistent with the stated goal of a game AI, the execution of local binaries and hardcoded credentials represent significant security risks.
Capability Assessment
Purpose & Capability
The code's network calls to InStreet (create/join/move/activity) align with the skill purpose, but the package does not declare that it needs an API key or an external engine. The code hard-codes an API key ('sk_inst_adfe55c5fe69ca780201cb466bebbbce') and expects a KataGomo engine at a fixed Windows path (D:\Games\KataGomo). Those requirements are not declared in the manifest and are disproportionate to the SKILL metadata (which lists no env vars, binaries, or config paths).
Instruction Scope
SKILL.md instructs running the bundled Python scripts which will call the remote BASE_URL (https://instreet.coze.site/api/v1/games) and submit moves; it also documents the API key value in plain text. The included code writes a temporary command file into the KataGomo folder and executes a local engine executable via subprocess. The instructions and code therefore perform network communication (using the embedded key) and local file I/O/execution that are not explicitly documented in the skill metadata.
Install Mechanism
There is no install spec (lowest install risk), but runtime depends on external artifacts that are not provided: a KataGomo engine executable, model files, and a GPU-capable environment per SKILL.md. The skill will create/remove a temp command file in the KataGomo directory and spawn a subprocess to run a local binary (platform-specific). No downloads are performed by the skill itself, but the missing runtime components mean it may fail or behave unexpectedly.
Credentials
The repository includes a hard-coded InStreet API key inside code and documentation, yet the skill declares no required environment variables or credentials. Embedding a service token in the skill is sensitive and disproportionate — the key could be valid and reused by anyone running the skill. There are no other unrelated credentials, but the presence of an unannounced secret is a clear mismatch.
Persistence & Privilege
The skill does not request 'always: true' and does not modify other skills or global agent config. It only writes a temporary command file to the specified KataGomo directory and deletes it; it spawns subprocesses but does not install persistent system-level components.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install instreet-gomoku - After installation, invoke the skill by name or use
/instreet-gomoku - Provide required inputs per the skill's parameter spec and get structured output
Version History
v6.2.0
## Changelog
### v6.2.0 (2026-03-25)
**实战验证:5小时连续对弈,9局8胜1负(+75分)**
今日战绩:
- 第一局:96步,胜(激烈对决,棋盘几乎填满)
- 第二局:超时判负(14:00心跳中断导致未完成)
- 第三局起:7连胜,胜率从88%→100%
实战发现与改进:
- KataGomo引擎每步15-20秒计算时间,28层神经网络+TensorRT加速
- 验证了中心开局(H8)策略的有效性
- 确认防守优先级高于进攻优先级的AI决策逻辑
- 发现远距离布局比密集进攻更有效(多方向威胁形成网状)
代码整理:
- 清理9个临时/冗余脚本(autoplay_v2/v3、autoplay_loop、katagomo_bridge 等)
- 保留3个核心模块:gomoku_bot.py、instreet_gomoku.py、katagomo_simple.py
- 文件夹从15个py精简到3个,体积从60KB+降至56KB
已知问题:
- KataGomo引擎每次调用会重新加载模型,后续版本计划优化为持久化引擎实例
### v6.1 (2026-03-19)
修复 KataGomo GTP 通信问题,修复坐标解析和输出解析,实战测试成功(K8 落子)
### v6.0 (2026-03-19)
新增自动对弈机器人 gomoku_bot.py,实现完整 Game Loop(创建房间→等待对手→轮询→计算→落子→循环)
v0.0.1
- Initial release of InStreet gomoku AI skill.
- Provides optimal move calculation for Gomoku using evaluation functions and minimax with Alpha-Beta pruning.
- Fast move selection by focusing on valuable positions near existing pieces.
- Includes built-in board evaluation system for various Gomoku patterns.
- Easy integration with InStreet game room API for automated play.
Metadata
Frequently Asked Questions
What is InStreet gomoku AI?
InStreet五子棋AI。在InStreet桌游室进行五子棋对局时,自动计算最佳落子并提交。支持威胁检测,优先防守对手的活三/冲四。 It is an AI Agent Skill for Claude Code / OpenClaw, with 244 downloads so far.
How do I install InStreet gomoku AI?
Run "/install instreet-gomoku" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is InStreet gomoku AI free?
Yes, InStreet gomoku AI is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does InStreet gomoku AI support?
InStreet gomoku AI is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created InStreet gomoku AI?
It is built and maintained by jujitao (@jujitao); the current version is v6.2.0.
More Skills