Token导航 LogoToken导航

Agent Skill · 研究检索

openword-navigator(打开词导航器)

openword-navigator 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

openword-navigator 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为接触敏感数据;许可证信息还不明确,第三方扫描有提醒项,按安装量在研究检索类的 73,675 个 Skill 里排第 69,749。

总安装 216周安装 9GitHub Stars 39下载 72核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:openword-navigator(打开词导航器)
来源仓库:https://github.com/dinghuanghao/openword
仓库路径:skills/openword-navigator
安装命令:
npx skills add https://github.com/dinghuanghao/openword --skill openword-navigator
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/dinghuanghao/openword --skill openword-navigator

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

openword-navigator 用于查找、检索和筛选相关信息。适用于 Codex、Claude、Cursor、Gemini CLI 中基于关键词或任务线索快速获取结果。通过 npx skills add 命令从指定 GitHub 仓库安装,具体能力以原始 README 为准。

安装前应核实权限、维护状态及是否触发网络访问或文件操作。适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

来源仓库内路径:skills/openword-navigator

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/dinghuanghao/openword --skill openword-navigat…;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要接触敏感数据,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

openword-navigator 怎么安装?

在终端执行 npx skills add https://github.com/dinghuanghao/openword --skill openword-navigator。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

openword-navigator 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

openword-navigator 需要什么权限?

权限级别是接触敏感数据。该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

openword-navigator 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

openword-navigator 用的人多吗?

公开数据是累计安装 216 次,近一周安装 9 次,下载 72 次,GitHub 39 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

openword-navigator 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。

先想清楚的地方

  • 需要接触敏感数据,装到常用环境前要先评估影响面。
  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 环境里有大量密钥、凭证,不方便给第三方 Skill 读取的人。
  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
openword-navigator

打开词导航器 · 本页

21639接触敏感数据未明确有提醒

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

OpenWord Player

Read references/session-runbook.md before running a session. Treat repository README.md as the source of truth for install/startup/API details, and re-check it when commands fail or look outdated.

Session Workflow

  1. Ask the user which mode to run before gameplay starts.
  2. Prepare runtime (download/install/start, browser, key setup).
  3. Branch to human-guided mode or AI REST mode.
  4. For every turn in both modes (including AI-driven sessions), show last_scene_image_path + narrative + player_profile as a natural user-facing update (avoid raw field-label dump).

Use this exact mode-selection question:

主人你想自己玩,还是我来玩?

If the user does not choose, default to human-guided mode.

Runtime Preparation

  1. If repository does not exist locally, run: git clone https://github.com/dinghuanghao/openword.git
  2. Install and start: npm install npm run dev Then check GET /health first: only if bridge_online=false, open one browser window at http://127.0.0.1:<PORT> (default 30000) and re-check once; if already connected, do not open a new window.
  3. GEMINI_API_KEY load order is: shell environment variable -> repo .env -> key modal/Settings.
  4. PORT is optional (default 30000); ensure browser opens http://127.0.0.1:<PORT>.
  5. Confirm the game can reach interactive state before sending turns.

Human-Guided Mode

  1. Confirm user can input actions in the browser.
  2. Help user craft opening world prompt and style.
  3. For each turn, provide a short recap and 3 action options with different risk/reward.
  4. Let the user pick or rewrite the action.
  5. Remind controls only when relevant: 机器人按钮 toggles built-in Auto Player, Esc exits auto mode/panels, Settings has Connect API Bridge.

AI REST Mode

Creative directive: in AI REST mode, be bold and imaginative; try seemingly impossible actions when they can produce a truly unique story.

  1. Ensure one browser tab is online at http://127.0.0.1:<PORT> (default 30000); before opening a new tab/window, check GET /health: if bridge_online=true, do not open; if bridge_online=false, open only one tab/window and re-check once.
  2. BFF/Connect API Bridge should be connected by default; do not ask the user to click it unless troubleshooting a bridge failure.
  3. Check server health (GET /health).
  4. Start or resume game via REST (create_game or load_game).

- If the player provides a reference image, always include it in create_game.image_path. - If the player does not provide a reference image, do not send image_path. - When image_path is used, prefer an absolute path.

  1. Loop: call get_current_game_state -> choose a high-impact next action -> call do_action -> immediately show a natural-language update synthesized from last_scene_image_path + narrative + player_profile to the user.
  2. After each turn update, either continue autonomously or pause at key nodes for user input.
  3. Prefer visual style 3D Pixel Art (voxel) or Claymation (clay); avoid frequent style switching in one session.

Use scripts/openword_rest.sh first (curl-only). If shell behavior is inconsistent across platforms, use scripts/openword_rest.py with the same commands/arguments. If PORT is not 30000, set: OPENWORD_BASE_URL="http://127.0.0.1:<PORT>".

REST API Schema

MethodPathBodySuccess fieldsTips
GET/healthnonestatus, bridge_onlineEnsure bridge_online=true before create_game, load_game, or do_action.
POST/api/create_game{"description": string, "style": string, "image_path"?: string}status, game_idIn description, cite the single closest reference game (tone + mechanics) for stable art direction and stronger worldbuilding; set style to voxel/clay first. If player provides a reference image, include image_path (absolute path preferred).
GET/api/show_history_gamesnonestatus, gamesUse after restart to find the most coherent checkpoint before loading.
POST/api/load_game{"game_id": string}statusAlways confirm selected game_id with the user before continuing.
GET/api/get_current_game_statenonestatus, game_id, world_view, narrative, player_profile, last_scene_image_pathCompare with previous state and extract one clear objective for the next turn.
POST/api/do_action{"description": string}status, game_id, world_view, narrative, player_profile, last_scene_image_pathPrefer interesting actions that directly advance the story; avoid low-impact micro actions that waste tokens.

Interaction Contract

Never run long silent streaks. Keep the user in the loop even when AI plays.

  1. After every turn, show one natural-language update based on last_scene_image_path + narrative + player_profile before the next action; do not print rigid prefixes like last_scene_image_path:, narrative:, player_profile:.
  2. At meaningful branch points, ask user preference before committing.
  3. Surface scene image paths and display scene images when possible.
  4. Highlight interesting moments: scene changes, risky decisions, major rewards, unexpected twists.

Failure Handling

  1. NO_BRIDGE: first verify the game tab is open and GET /health status; only then ask user to enable Connect API Bridge in Settings for troubleshooting.
  2. Bridge occupied: another tab owns the bridge; disconnect that tab first.
  3. Missing key/model errors: check GEMINI_API_KEY in shell env first, then repo .env; if still missing, ask user to configure key modal/Settings.
  4. Slow/timeout calls: increase timeout, avoid overlapping requests. Some APIs may return BUSY; wait patiently and retry only after the current call settles, which can occasionally take several minutes when the service is blocked.

平台分布

Codex

37.13%
按下载量换算27

Claude

28.16%
按下载量换算20

Cursor

18.27%
按下载量换算13

Gemini CLI

9.26%
按下载量换算7

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

可疑

权限和风险

接触敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在提醒或未通过项,安装前请自行复核。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills