The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions in SKILL.md and the environment check script scripts/check-permissions.ts suggest installing the Bun runtime using the command curl -fsSL https://bun.sh/install | bash. This targets the official distribution site for a well-known developer tool.
[EXTERNAL_DOWNLOADS]: The conversion script scripts/md-to-wechat.ts downloads remote images specified in user-provided markdown files using the https and http modules for local processing.
[EXTERNAL_DOWNLOADS]: The syntax highlighting utility scripts/md/utils/languages.ts dynamically imports language grammars from an external CDN (cdn-doocs.oss-cn-shenzhen.aliyuncs.com) at runtime.
[COMMAND_EXECUTION]: The skill uses node:child_process utilities (spawn, spawnSync, execSync) across multiple scripts to launch Google Chrome with a dedicated debugging port and profile directory, execute internal rendering and clipboard management scripts, and trigger system-level keystrokes (osascript on macOS, xdotool on Linux, PowerShell on Windows) to simulate pasting in the WeChat editor.
[CREDENTIALS_UNSAFE]: The skill manages WeChat API credentials (WECHAT_APP_ID, WECHAT_APP_SECRET). It prompts users to store these secrets in a .env file within the .baoyu-skills/ directory. scripts/wechat-api.ts reads these credentials to obtain access tokens from the official WeChat API endpoint (api.weixin.qq.com).
[PROMPT_INJECTION]: The skill processes user-provided markdown files to extract metadata such as titles and summaries (scripts/md-to-wechat.ts, scripts/wechat-browser.ts). This serves as an indirect prompt injection surface where malicious instructions within a processed file could potentially target the agent's logic during the article preparation workflow.

baoyu-post-to-wechat