Spot it. Save it. Free, local-first Chrome extension for images, videos and prompts. Explore 2,000+ curated cases and connect your library to AI Agents via MCP. 提示词导演
Distribution and permissions
Public distribution formats and the permissions used by PromptDirector.
Stable Web Store identity
- Chrome Web Store item public key: stored in the source
manifest.jsonfor local identity verification only;package:releaseremoveskeyfrom the upload manifest. - Expected extension ID:
iahakaahijddcjjldidbclicedibgpjm. - Current upload artifact:
dist/store/PromptDirector-1.28.0-Chrome-Web-Store.zip. Keep this artifact local for Web Store submission. The sole GitHub extension installation package isPromptDirector-1.28.0.zip, with stable identity and the local updater. - Generate a candidate only with
npm run package:release; the command verifies the source public key against the expected identity, then removeskeyfrom the Web Store upload manifest and excludes the local program installer.
Store listing
- Chinese copy:
store/LISTING_ZH_CN.md - English copy:
store/LISTING_EN.md - Public privacy policy URL:
https://promptdirector.cn/privacy.html - Homepage URL:
https://promptdirector.cn/ - Support URL:
https://promptdirector.cn/support.html - 128x128 icon:
assets/icons/icon-128.png - 440x280 small promo tile:
store/small-promo-440x280.png - Screenshots, in order:
store/screenshots/01-library-1280x800.pngstore/screenshots/02-skills-1280x800.pngstore/screenshots/03-composer-1280x800.png
Single purpose
Help visual creators capture web visuals with their prompts, media, and notes as a searchable, reusable local creative archive.
中文用途说明:
帮助视觉创作者在用户主动操作下,将网页画面、提示词、媒体和笔记保存为可检索、可复用的本地创作资料,并以这些资料创建创作 Skill 和提示词。
Permission justifications
activeTab: read only the page on which the user explicitly starts capture, visual selection, or a screenshot action.alarms: schedule, poll, pause, and resume local background jobs without tracking browsing.contextMenus: expose the user-triggered right-click capture command.downloads: save backups, share packages, or media files requested by the user.offscreen: process screenshots, media, documents, and ZIP archives locally in a short-lived background document.sidePanel: provide the capture and creation workspace in Chrome's side panel.scripting: inject bounded capture or selection code only after the user starts that operation.storage: store the user's local cases, projects, settings, and task state.unlimitedStorage: retain the images, videos, and documents the user explicitly saves to the local archive.https://wchao6891.github.io/*: read the official PromptDirector curated catalog, metrics, covers, and previews.https://github.com/*: read a curated GitHub Release selected by the user and open public feedback or submission pages.https://objects.githubusercontent.com/*: follow GitHub's Release download path for selected curated content.https://release-assets.githubusercontent.com/*: download selected curated Release archives and media after integrity checks.- Optional
<all_urls>: not granted at installation. Capture and AI settings request the necessary origins. Enabling Agent connection requests website access so the paired Agent can capture user-requested URLs in a dedicated tab; this does not monitor browsing. - Optional
downloads.open: requested when the user installs the video component; the explicit Open installer action opens the downloaded official installer. Files are never opened automatically. - Optional
nativeMessaging: requested when enabling Agent connection or installing the video component. Agent connection supports case search, original-file access, requested capture and material saving. Independent local video extraction resolves and merges selected video originals without Agent pairing. - Optional
clipboardRead: read clipboard content only when the user triggers the relevant action. - Optional
declarativeNetRequestWithHostAccess: create bounded media-read rules only on origins the user already granted.
Chinese permission justifications
- Host permissions: 固定的 GitHub Pages 与 GitHub 域名仅用于读取官方精选目录、封面、审核记录及用户选择下载的精选 Release 包。其他网站和 AI 服务使用可选
<all_urls>,普通采集和服务配置请求具体来源;用户启用 Agent 连接时申请网站访问,以便按指令在独立标签页采集网址。不在安装时授予,不被动监控浏览。 activeTab: 用户点击扩展、右键菜单、选图或截图后,临时读取当前标签页中的用户可见内容并完成该次采集。不会被动访问其他标签页或后台记录浏览。alarms: 用于安排、轮询、暂停和恢复用户已启动的本地后台任务,使服务工作器休眠后仍可继续;不用于追踪浏览活动。clipboardRead: 仅当用户点击从剪贴板提取文字或图片的明确动作时,请求可选权限并读取本次剪贴板内容;不在后台持续读取。contextMenus: 在网页右键菜单提供用户主动触发的收藏入口,将选中文字或当前页面内容送入本地采集流程。declarativeNetRequestWithHostAccess: 仅在用户已授予具体站点访问权限、且保存已确认媒体需要调整该次资源请求时,创建范围受限的临时规则;不拦截或修改无关网站流量。downloads: 将用户明确要求的资料备份、分享包或媒体文件保存到本机下载目录;不会自动下载未知文件。offscreen: 在短期离屏文档中完成需要 DOM 或 Canvas 的本地处理,例如截图裁剪、缩略图、媒体或文档解析和 ZIP 处理;不显示广告、不追踪用户。scripting: 仅在用户主动启动网页采集、智能选图或截图时,向当前页面注入随扩展打包的采集或选择脚本;不执行远程代码,不对未授权页面持续注入。sidePanel: 在 Chrome 侧边栏提供采集、整理和创作工作区,使用户在当前网页旁预览并确认要保存的内容。storage: 在浏览器扩展本地存储中保存用户的案例、项目、标签、设置、任务状态和用户配置的 AI 服务信息;默认不上传开发者服务器。unlimitedStorage: 用户可能主动保存大量图片、视频和文档,标准配额不足以可靠保留完整本地创作资料库;权限只用于用户选择保存的本地内容。
Privacy practices
- Policy reference: Manifest V3 requirements require packaged executable logic; the upload build excludes the GitHub program installer.
- Distribution updates: the Web Store archive contains no local program installer; updates are delivered by Chrome. The separate GitHub development archive supports user-authorized in-place program updates.
- Remote code: No, the Web Store extension does not use remote code. Remote JSON, media, curated archives, and AI responses are treated as data and are never executed as extension code.
- User data: disclose website content and resources, saved source URLs/browsing activity, user-provided prompts/notes/media, and locally stored AI-service authentication information. These categories are handled even when they remain on-device.
- Data transfer: user-selected content is sent only to the AI service or compatible endpoint selected and authorized by the user. Curated catalog and Release requests contain no private library data.
- Limited Use: certify all applicable statements. There is no sale of user data, personalized advertising, analytics collection, or developer-operated content server.
Chinese remote-code declaration:
商店包不包含本地程序安装器,更新由 Chrome 分发。不使用远程代码。所有可执行 JavaScript、CSS、PDF 与文档解析器和 ZIP 逻辑均随扩展包提供。远程 JSON、媒体、精选包和 AI 响应只作为经过校验的数据处理,绝不作为代码执行。
Data types handled, including local-only handling:
- Authentication information: user-configured AI API keys and sending authorization.
- Web browsing activity: source URLs and origins involved in a user-triggered capture.
- Website content and resources: user-selected text, images, videos, documents, and screenshots.
- User-generated content: prompts, notes, cases, projects, tags, Skills, and creative requests.
See the privacy policy for the data handled by user-triggered operations.
Release artifacts
Run npm run package:distribution to build both channels. Publish only the versioned directory under dist/github/: it contains the user installation ZIP and SHA256SUMS. The Web Store upload ZIP stays under dist/store/ for submission. Agent connector assets use the dedicated agent-connector-v<connector-version> release, marked as not latest, and are linked from connector/INSTALL.md.
webRequest: Observe media response URLs and content types on already authorized sites to populate tab-specific video badges and capture candidates. No request modification, credential collection or browsing-history upload. Cross-origin media still requires host access.