The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill dynamically imports JavaScript code at runtime from a remote CDN (cdn-doocs.oss-cn-shenzhen.aliyuncs.com) to provide syntax highlighting for different programming languages. This creates a vector where unverified remote code is executed within the agent's environment.
[COMMAND_EXECUTION]: Multiple scripts (paste-from-clipboard.ts, copy-to-clipboard.ts, check-permissions.ts) execute system-level commands across different platforms. This includes osascript and swift on macOS, powershell.exe on Windows, and xdotool or ydotool on Linux. These are used for clipboard automation and environment verification.
[DATA_EXFILTRATION]: The skill is designed to read sensitive credential files, specifically .env and ~/.super-publisher/.env, to retrieve WeChat API credentials (app_id and app_secret). While necessary for the skill's purpose, this pattern involves the handling and transmission of secrets to external WeChat APIs.
[EXTERNAL_DOWNLOADS]: The skill downloads remote images referenced in Markdown files and fetches content from external services like plantuml.com and a custom CDN for syntax highlighting logic.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a data pipeline that ingests local Markdown files and transforms them into HTML for injection into a browser-based editor. This ingestion surface could potentially be used to inject malicious HTML or instructions into the target web application if the source files are not controlled by the user.

wechat-post-publisher