PaperCash — 论文全流程辅助 Skill
ReviewAudited by ClawScan on May 10, 2026.
Overview
PaperCash is a coherent academic-writing helper, but it deserves review because it explicitly offers AI-detection evasion and asks users to store academic-site browser cookies.
Install only if you are comfortable running a local Python research tool that contacts academic websites. Be cautious with the “降AI率” feature, avoid using it to misrepresent authorship, and do not paste CNKI/Wanfang browser cookies unless you understand that they are sensitive account-session data. Use a virtual environment and avoid checking confidential drafts unless you accept that snippets may be sent to external search services.
Findings (5)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
Using this feature could violate school or publisher rules and may expose the user to academic integrity consequences.
The skill explicitly instructs the agent to transform text to reduce AI-generated-content detection, which can be used to misrepresent authorship or bypass academic review processes.
### 5. 降AI率改写 (humanize) 通过句式变换、个人观点注入、学术语气调整，降低AI生成内容的检测率。
Use the tool only for legitimate editing and disclose AI assistance where required; the publisher should reframe this feature as transparent writing improvement rather than detector evasion.
If those cookies are exposed through local file access, logs, backups, or misuse by the skill, someone could potentially access the user’s CNKI/Wanfang session.
The setup flow asks users to copy browser cookies into a local config file. Browser cookies are sensitive session credentials that may grant account access beyond simple search.
# 获取: 浏览器登录知网 -> F12 -> Network -> 复制 Cookie # CNKI_COOKIE= # 万方 Cookie（可选） # WANFANG_COOKIE=
Avoid pasting full browser cookies unless necessary; declare these credentials in metadata, store them securely, document exactly how they are used, and warn users to remove or rotate them when done.
Private or unpublished draft text may be transmitted as search queries to academic data providers.
The plagiarism pre-check is designed to send sentence-level content from a user’s paper to external academic search sources. This is aligned with the feature, but it creates a third-party data boundary.
查重预检 (check) 将论文拆成句子，逐句检索学术库，标记高相似度句子。
Do not use confidential drafts unless you are comfortable with snippets being sent to external services; the skill should add an explicit privacy warning before check operations.
Other local processes or backups may retain evidence of research topics and cached results.
The skill stores cached responses under the user’s home directory for up to 24 hours, which can preserve research queries and retrieved paper metadata locally.
_CACHE_DIR = Path.home() / ".cache" / "papercash" _DEFAULT_TTL = 3600 * 24 # 24 hours
Clear the cache for sensitive topics and consider documenting the cache location, retention period, and cleanup command.
A future dependency update could change behavior or introduce vulnerabilities.
The package dependencies use minimum versions rather than pinned exact versions. This is common for Python tools, but it leaves installed code dependent on future package resolution.
jieba>=0.42.1 requests>=2.31.0 beautifulsoup4>=4.12.0 python-docx>=1.1.0
Install in a virtual environment and prefer a reviewed lockfile or pinned dependency versions for reproducible installs.