The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands using subprocess.run to interact with lark-cli for fetching document data and ffmpeg for video processing.
Evidence: scripts/wechat_draft_publisher/lark_docs.py and scripts/wechat_draft_publisher/wechat_api.py call lark-cli and ffmpeg with arguments derived from input parameters.
[EXTERNAL_DOWNLOADS]: The skill downloads content from external sources, including Feishu documents, images, and video assets.
Evidence: scripts/wechat_draft_publisher/wechat_api.py uses urllib.request.urlopen to communicate with api.weixin.qq.com and open.feishu.cn.
Evidence: scripts/wechat_draft_publisher/pipeline.py downloads remote images for article covers.
[PROMPT_INJECTION]: The skill processes untrusted content from Feishu documents which could potentially contain malicious instructions (indirect prompt injection) meant to influence an agent reading the rendered output.
Ingestion points: scripts/wechat_draft_publisher/lark_docs.py (via lark-cli output).
Boundary markers: Absent. Content is processed and rendered into HTML without explicit delimiters for the agent context.
Capability inventory: Shell command execution via lark-cli and ffmpeg across multiple scripts, network requests to WeChat/Feishu APIs, and local file writing for previews and caching.
Sanitization: Employs HTML escaping and regex-based normalization for specific Markdown elements.

feishu-doc-to-wechat-draft