The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: Fetches external source code from a GitHub repository (wechat-article/wechat-article-exporter) pinned to a specific commit hash (6b67dfe64f6f359be604239e98f74c1021fc9d5f) and applies a local privacy patch during the Docker backend setup.
[REMOTE_CODE_EXECUTION]: Installs Python dependencies (Selenium, BeautifulSoup4, markdownify) from the official package registry using a hash-locked requirements file (assets/lite-requirements.lock) to ensure package integrity during the Lite backend installation.
[COMMAND_EXECUTION]: Invokes shell commands via subprocess.run to manage Git operations and Docker Compose services for the archive backend lifecycle.
[COMMAND_EXECUTION]: Employs os.execve to re-execute the skill within a dedicated virtual environment, ensuring the Lite backend runs with isolated dependencies.
[PROMPT_INJECTION]: The skill processes untrusted article content and metadata from mp.weixin.qq.com, representing an attack surface for indirect prompt injection.
Ingestion points: scripts/wechat_mp_exporter.py and scripts/lite_backend.py retrieve HTML, text, and Markdown content from external article URLs.
Boundary markers: The skill lacks explicit prompt-level delimiters or instructions to isolate the fetched article content from the agent's primary instructions.
Capability inventory: The skill possesses file-writing capabilities (for local archiving) and command execution capabilities (Git, Docker, and Python subprocesses).
Sanitization: Content is filtered to specific HTML elements (#js_content), but the resulting text is not sanitized to remove potential natural language instructions targeting the LLM.

wechat-mp-exporter