The Agent Skills Directory

[REMOTE_CODE_EXECUTION] (HIGH): The skill dynamically downloads and executes JavaScript from an untrusted remote CDN, which could allow for arbitrary code execution if the CDN is compromised.\n
Evidence: In scripts/md/utils/languages.ts, the loadAndRegisterLanguage function uses the import() function to load code from https://cdn-doocs.oss-cn-shenzhen.aliyuncs.com/npm/highlightjs/....\n- [DATA_EXFILTRATION] (HIGH): The skill can be coerced into reading and exposing sensitive local files by processing them as markdown image attachments.\n
Evidence: In scripts/md-to-wechat.ts, the resolveImagePath function resolves absolute local file paths provided in the markdown source. These files are then read and prepared for publication (pasted via the clipboard) in the WeChat editor.\n- [COMMAND_EXECUTION] (MEDIUM): The skill executes platform-specific shell commands to control the host system and simulate user keystrokes.\n
Evidence: scripts/paste-from-clipboard.ts executes osascript (macOS), powershell.exe (Windows), and xdotool (Linux). scripts/md-to-wechat.ts executes npx -y bun to run rendering scripts.\n- [EXTERNAL_DOWNLOADS] (LOW): The skill automatically fetches remote image assets found in markdown content.\n
Evidence: scripts/md-to-wechat.ts uses a downloadFile function to fetch content from arbitrary URLs.

baoyu-post-to-wechat