菜单
Read and analyze PDF files, extract text by page range, and produce structured summaries with citations. Use when a user asks to read, summarize, compare, or QA against PDF documents, including long reports, papers, contracts, manuals, and scanned PDFs.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Operate the current Skuare CLI end to end in the current workspace. Use when the task is to initialize `skr`/`skuare`, manage remote sources, inspect or install Skills/AGENTS.md, remove local installs, publish or migrate registries, or update local skill dependency files.
Coordinate ingestion, reporting, and alert routing into a full observability workflow with end-to-end traceability.
Route ingestion anomalies to downstream channels with severity policies, schedule windows, and deduplication controls.
Diagnose API integration failures by tracing request/response details, auth, and environment differences.
Fetch upstream API payloads, normalize records, validate schema, and output ingestion-ready batches.
Provide reusable HTTP request execution patterns including retries, timeout boundaries, and response sanity checks.
| name | pdf-reader |
| metadata | {"version":"0.0.1","author":"ShuangShu"} |
| description | Read and analyze PDF files, extract text by page range, and produce structured summaries with citations. Use when a user asks to read, summarize, compare, or QA against PDF documents, including long reports, papers, contracts, manuals, and scanned PDFs. |
Use this skill to read PDF content safely and reproducibly, then return concise answers with page-level evidence.
Use the fastest available method first, then fallback:
pdftotext (if installed) for quick extraction:pdftotext -layout input.pdf -
pypdf for controlled page-by-page extraction:from pypdf import PdfReader
reader = PdfReader("input.pdf")
for i, page in enumerate(reader.pages, start=1):
text = page.extract_text() or ""
print(f"\\n--- Page {i} ---\\n{text}")
ocrmypdf + extraction), then clearly mark OCR uncertainty.Always return:
结论: 直接回答用户问题。依据: 关键证据点，附页码（如 p.12）。不确定性: 缺失页、识别错误、扫描噪声等风险。后续建议: 仅在确有必要时给 1-2 条下一步。