The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses various system-level commands to automate browser interactions and clipboard operations across different platforms.
Evidence: Uses osascript on macOS, powershell.exe on Windows, and xdotool/ydotool/xclip/wl-copy on Linux to simulate keystrokes (Cmd+V/Ctrl+V) and manage the system clipboard for pasting HTML and images.
[EXTERNAL_DOWNLOADS]: Fetches required components and assets from external services at runtime.
Evidence: Downloads code syntax highlighting definitions from the Doocs project's Alibaba OSS CDN (cdn-doocs.oss-cn-shenzhen.aliyuncs.com) and renders diagrams via the official PlantUML service.
[REMOTE_CODE_EXECUTION]: Dynamically imports scripts for syntax highlighting from a remote CDN.
Evidence: scripts/md/utils/languages.ts uses dynamic import() to load highlight.js language modules from a well-known third-party CDN at runtime.
[PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it processes user-provided markdown content that influences agent behavior.
Ingestion points: Markdown or plain text files provided by the user (e.g., scripts/wechat-article.ts).
Boundary markers: No specific delimiters or "ignore" instructions are used to wrap the untrusted content.
Capability inventory: Subprocess execution via spawnSync, network requests via fetch, and full browser control via Chrome DevTools Protocol (CDP).
Sanitization: Content is parsed using standard markdown libraries without additional sanitization or instruction filtering.

baoyu-post-to-wechat