下面是一组仅面向“PDF工具”类别的在线工具生成型 AI 提示词（Prompts）。每条 Prompt 都要求 AI 直接产出可运行/可构建/可部署的在线 PDF 工具：给出完整项目代码、文件结构、运行命令、部署说明，以及不少于 5 条测试用例或 QA checklist。你可以按需挑选其中一条，让 AI 一次性生成一个独立可用的小工具。
在线 PDF 对比与差异高亮（文本 + 页面层级）工具
用于对比同一份文档的两个版本，支持逐页渲染预览与差异高亮导出报告。
英文 Prompt:
You are a senior full-stack engineer. Build a production-ready web app: “Online PDF Compare & Diff Highlighter”.
Requirements:
- Users upload two PDF files (A and B) in the browser.
- Render page thumbnails and a side-by-side page viewer using pdf.js.
- Extract text per page, compute diffs (word-level and line-level). Provide a toggle to switch diff granularity.
- Highlight insertions/deletions/changes in the viewer (overlay on top of rendered pages).
- Generate a downloadable diff report: JSON + a human-readable HTML report.
- Handle large PDFs (up to 200 pages) with streaming/worker- d parsing and progress UI.
Tech stack:
- Next.js (App Router) + Type .
- pdf.js for rendering and text extraction.
- Web Workers for heavy parsing/diff computation.
- No external paid APIs.
Deliverables (must output all):
1) Full file tree.
2) Complete source code for every file.
3) Local run commands.
4) Dockerfile + docker-compose.
5) Deployment guide (Vercel + self-host).
6) At least 8 tests or QA checklist items (include edge cases like scanned PDFs, missing fonts, rotated pages).
Implementation notes:
- Provide a simple diff algorithm (e.g., Myers or patience diff) and explain complexity.
- Keep everything client-side if feasible; if server-side is required, justify and keep uploads ephemeral.
中文释义:生成一个可直接部署的在线 PDF 版本对比工具：上传两份 PDF，逐页预览并高亮差异，同时导出对比报告，适合合同/标书/论文的版本审阅。
在线 PDF 水印检测与批量移除工具（仅对已有水印做处理）
用于识别并移除重复出现的文本水印/图形水印，并提供操作预览与回滚。
英文 Prompt:
Build a web app: “PDF Watermark Detector & Remover (Preview + Batch)”.
Constraints:
- This tool must NOT generate new images. Only detect/remove watermarks that already exist in the uploaded PDF.
Features:
- Upload one or multiple PDFs.
- Detect repeated watermark patterns (text overlays, low-opacity stamps) using page inspection.
- Let the user preview before/after per page.
- Provide 3 removal strategies: hide s, replace content stream segments, and rasterize page as last resort (warn about quality).
- Export cleaned PDFs and a processing log.
Tech stack:
- React + Type .
- Use a JS PDF manipulation library (pdf-lib or similar) and explain limitations.
Deliverables:
- Full project code, file tree, commands.
- Deployment instructions.
- At least 6 QA checklist items (encrypted PDF, permissions, mixed rotations, huge files, CJK fonts).
中文释义:让 AI 生成一个在线 PDF 水印检测/移除工具，只对现有水印做处理，支持批量处理、预览对比与导出处理日志，适用于资料清理与合规审阅前的文档整理。
在线 PDF 目录（书签）自动生成与修复工具
用于为没有书签的 PDF 自动生成目录结构，或修复错误层级/重复标题。
英文 Prompt:
Create a web app: “PDF Outline (Bookmarks) Generator & Fixer”.
Requirements:
- Upload a PDF.
- Extract headings heuristically from text: font size, boldness, numbering patterns, indentation.
- Provide an interactive outline editor (drag to reorder, change levels, rename, merge).
- Write the outline back into the PDF as bookmarks.
- Export the updated PDF and an outline JSON.
Stack:
- Next.js + Type .
- pdf.js for extraction + pdf-lib for writing outlines (if writing outlines is limited, provide a fallback: generate a sidecar outline file and explain).
Deliverables:
- Full code + file tree + run/deploy instructions.
- Provide at least 5 test PDFs scenarios and expected results.
中文释义:生成一个在线“PDF 书签/目录生成器”，自动从标题结构推断目录并允许手动调整，最后回写到 PDF 里，适合长文档交付与阅读导航优化。
在线 PDF 批注汇总与审阅意见导出工具
用于提取批注、高亮、便签等审阅信息，统一导出为可追踪的任务清单。
英文 Prompt:
Build “PDF Annotation Extractor & Review Exporter”.
Features:
- Upload a PDF that contains annotations (highlights, sticky notes, underline, strikeout).
- Parse annotations per page with coordinates.
- Show an annotations list with filters (author, type, page).
- Export to CSV and Markdown, including deep s like “Page 12, highlight text: …”.
- Optionally generate a simple reviewer checklist view.
Tech:
- Type + pdf.js.
- Keep processing local in the browser.
Deliverables:
- Full project code, file tree.
- Commands to run.
- Deployment instructions.
- QA checklist (at least 8 items) including multiple authors, rotated pages, and missing annotation data.
中文释义:生成一个在线“PDF 批注提取器”，把高亮/便签等审阅意见集中成列表并导出为 CSV/Markdown，便于协作流转与整改闭环。
在线 PDF 元数据与隐私信息清理工具（含预览与差异报告）
用于查看并清理作者、软件信息、隐藏字段、XMP 等元数据，输出清理前后差异。
英文 Prompt:
Create “PDF data & Privacy Scrubber”.
Requirements:
- Upload a PDF.
- Display all detectable data (Info dict, XMP, embedded file attachments list, form field names).
- Let user choose what to remove/normalize (author/ /producer/creation dates/XMP/custom keys).
- Output a cleaned PDF.
- Also output a diff report showing what changed.
Stack:
- React + Type .
- Use pdf-lib for rewriting data; explain limitations.
Deliverables:
- Full project code and file tree.
- Run and deploy instructions.
- At least 6 QA items: encrypted, signed PDFs (warn), incremental updates, attachments.
中文释义:生成一个在线“PDF 隐私清理”工具：可视化展示元数据并可选择性清理，同时导出差异报告，适合对外分享前的信息脱敏。
在线 PDF 表格抽取与结构化导出工具（不生成新图片）
用于从 PDF 中抽取表格为结构化数据，导出 CSV/JSON 并提供抽取准确性校验。
英文 Prompt:
Build “PDF Table Extractor to CSV/JSON (Validation UI)”.
Constraints:
- Do not generate images. Only process existing PDF content.
Features:
- Upload PDF.
- Detect tables using text coordinates and ruling lines heuristics.
- Provide a UI to preview extracted tables and edit cells.
- Export per-table CSV and a combined JSON.
- Provide confidence indicators and a validation checklist.
Tech:
- Next.js + Type .
- pdf.js for text positions.
Deliverables:
- Full code + file tree + run commands.
- Docker + deploy guide.
- At least 8 QA items including multi-line cells, merged cells, rotated tables.
中文释义:生成一个在线“PDF 表格抽取器”，从已有 PDF 内容中提取表格并导出 CSV/JSON，带可视化校验与手工修正能力，适合财务报表/清单/报价单结构化。
在线 PDF 分页拆分与命名规则批处理工具（可预览与回滚）
用于按页范围拆分、按正则从首页文字提取命名、批量导出 Zip。
英文 Prompt:
Create “PDF Splitter + Naming Rules Batch Export (Preview + Undo)”.
Features:
- Upload a PDF.
- Define split rules: fixed page ranges, every N pages, or detect separators by keyword on page.
- Define naming rules: regex on extracted text (e.g., invoice number) with fallback numbering.
- Preview the output list before export.
- Export as individual PDFs and optionally as a ZIP.
Tech:
- React + Type .
- pdf-lib for splitting.
Deliverables:
- Full code + file tree + commands.
- Deployment instructions.
- QA checklist (at least 6) for edge cases: empty pages, huge PDFs, unicode filenames.
中文释义:生成一个在线“PDF 拆分批处理工具”，支持拆分规则+命名规则的预览与回滚，适合票据/合同批量分发与归档。
在线 PDF 可访问性（A11y）检查与问题清单导出工具
用于检查标签树、标题结构、阅读顺序、替代文本等，并生成整改清单。
英文 Prompt:
Build “PDF Accessibility (A11y) Checker & Issues Exporter”.
Features:
- Upload a PDF.
- Run checks: presence of tags, heading structure, document language, reading order hints, embedded fonts, form field labels.
- Output an issues list with severity, remediation suggestions, and references to PDF sections/pages.
- Export issues as JSON and Markdown.
Tech:
- Next.js + Type .
- If full tag parsing is not feasible in browser, implement partial checks and clearly document limitations.
Deliverables:
- Full project code, file tree.
- Commands to run.
- Deployment guide.
- At least 8 QA items (tagged vs untagged PDFs, CJK, scanned-only PDFs).
中文释义:生成一个在线“PDF 可访问性检查器”，输出问题清单与整改建议并支持导出，适合对外发布前的合规自查与团队协作整改。
