wechat-style-writer
Security checks across malware telemetry and agentic risk
Overview
This writing skill is mostly scoped to rewriting articles, but it explicitly tells the agent to make AI-written text look human by adding deliberate mistakes, which could mislead readers.
Install only if you are comfortable with a style-rewriting tool that closely imitates named writers. Before publishing, remove the deceptive typo instruction, disclose AI assistance where appropriate, and avoid implying the original author wrote or endorsed the article.
VirusTotal
VirusTotal findings are pending for this skill version.
Risk analysis
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.
Published output could mislead readers into thinking the article was written naturally by a human rather than generated or edited by AI.
The skill explicitly instructs the agent to make AI-generated writing appear human by inserting intentional typos and imperfections.
生成的文章需要模拟真人赶稿时的"不完美感"，让文章看起来像博主深夜快速敲出来发的，而不是AI逐字精雕细琢的。具体做法：在全文中随机制造1-2个常见错别字...
Remove instructions that intentionally disguise AI authorship. If naturalness is desired, improve tone and flow without adding deceptive mistakes, and disclose AI assistance when publishing.
Users may still publish content that readers associate with the named authors unless they clearly disclose that it is only style-inspired.
The skill recognizes impersonation risk and includes safeguards to remove personal identifiers, but it still centers on closely mimicking named authors' styles.
改写是模仿写作风格，不是冒充原作者。必须移除所有与原作者个人身份绑定的内容...
Keep the anti-impersonation safeguards, avoid using protected names as bylines, and clearly label generated work as style-inspired rather than authored by the referenced person.
The agent may create local output files when the user agrees.
The skill can write an output Markdown file and present it for download, but only after explicit user confirmation and to a scoped output directory.
如果用户确认保存，再将文件保存到 `/mnt/user-data/outputs/` 并使用 `present_files` 工具提供下载
Review the generated article before confirming save, and ensure the filename and content are acceptable.
A poorly curated or maliciously edited style profile could influence generated articles in unexpected ways.
Persistent style files and user-added examples become influential context for future generations; if a style file contains unwanted instructions, the agent may over-weight them as reference material.
精选范文是最直接的风格参照，优先级高于文字描述... 文件放到 `references/styles/` 目录即可生效
Only add style profiles from trusted sources, review them for embedded instructions or private information, and treat examples as data rather than commands.