Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
paper-reader (XuRuitian version)
v1.0.0精读学术文献的专家级 Skill。当用户上传 PDF、Word、Excel、PPT 或 TXT 格式的学术论文，并希望进行深度学术分析时使用本 Skill。支持中英双语文献， 可自动识别文件类型、提取全文内容，并按六大维度（研究目标、新方法、实验验证、 未来方向、批判分析、实用建议）输出结构化分析报告。触发词包括...
⭐ 0· 67·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
The skill name/description (paper reader that extracts text and generates a report) aligns with the included scripts: a Python extractor for multiple document formats and a Node.js report generator. Required dependencies (Python libs, docx) are proportional to the stated functionality.
Instruction Scope
SKILL.md instructs the agent to extract text from the uploaded file, follow the provided prompt template, output a Markdown report, and then produce a Word document saved to the user's Desktop and write JSON to <skill_path>/data/latest_analysis.json. These actions are consistent with the purpose, but the requirement to always save to the user's Desktop and to persist latest_analysis.json to the skill directory are operational details the user should be aware of (they create local files outside the skill folder). The references file also suggests optionally using external searches (Tavily) for up-to-date literature, which would involve network access if the agent follows it — but the skill's scripts themselves make no network calls.
Install Mechanism
There is no automated install specification; the README instructs manual installation of standard Python and Node.js packages (PyMuPDF, pdfplumber, python-docx, openpyxl, docx npm package). This is a normal approach for a script-based, local-processing skill and does not pull arbitrary remote archives or run unexpected installers.
Credentials
The skill declares no required environment variables, no credentials, and no config paths. The code does not read environment variables or request secrets. Calls to system tools (textutil on macOS, antiword on Windows) are optional fallbacks for .doc support and are documented.
Persistence & Privilege
The skill does write output files: it recommends saving JSON to <skill_path>/data/latest_analysis.json and saving the generated .docx to the user's Desktop. 'always' is false and the skill does not attempt to modify other skills or system-wide config. Users should note the persistent storage of analysis results in the skill data folder and the automatic Desktop write behavior.
Assessment
This skill appears to do what it says: extract text from uploaded documents, produce a structured analysis following the provided template, and generate a Word report. Before installing or running it: 1) Review and be comfortable with local file writes — the skill will create/overwrite data/latest_analysis.json in the skill folder and will save a .docx to your Desktop (SKILL.md requires this). 2) Install the documented Python and Node dependencies in a controlled environment; the Node script requires the 'docx' package. 3) The extractor can invoke local system tools (textutil or antiword) as fallbacks for .doc files — those commands run locally and require those tools to be present. 4) The prompt reference suggests optional web searches for up-to-date literature; the provided scripts do not perform network requests, but if you (or an agent) follow that suggestion the agent may use the network — only allow that if you trust the agent's browsing behavior. 5) The README claims a security audit; treat that as an unauthenticated claim unless you have audit artifacts. If you need stronger guarantees, run the scripts in a sandbox or inspect/execute them manually on non-sensitive files first.Like a lobster shell, security has layers — review code before you run it.
academicvk974ey0qmtep6n4aka7njhp0hn83vs4tlatestvk974ey0qmtep6n4aka7njhp0hn83vs4tpdfvk974ey0qmtep6n4aka7njhp0hn83vs4tresearchvk974ey0qmtep6n4aka7njhp0hn83vs4t
License
MIT-0
Free to use, modify, and redistribute. No attribution required.