The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/doctor.py tool uses subprocess.run to execute the macOS networksetup utility. This is strictly used to retrieve lists of network services and proxy statuses for environment validation, and does not perform any system modifications.\n- [COMMAND_EXECUTION]: The scripts/start_wxdown_service.py script employs subprocess.call to launch a local Python-based helper service (wxdown-service). The execution path is derived from user-specified directories or environment variables, intended for interacting with local developer tools.\n- [EXTERNAL_DOWNLOADS]: The scripts/download_urls.py and scripts/doctor.py scripts make HTTP GET requests to https://down.mptext.top using the standard urllib.request library. This external domain provides the primary API for downloading and converting WeChat articles into Markdown or JSON formats.\n- [PROMPT_INJECTION]: The skill processes external article content and metadata from the WeChat platform, which represents an indirect prompt injection surface (Category 8). Ingestion points: article body fetching in scripts/download_urls.py and history JSON reading in scripts/analyze_history.py. Boundary markers: The skill contains explicit instructional blocks in SKILL.md and references/manual-gates.md requiring user confirmation and prohibiting autonomous execution. Capability inventory: local command execution via subprocess in diagnostic and helper scripts. Sanitization: While explicit code-level sanitization for LLM instructions is absent, the skill's rigid manual confirmation gates mitigate the risk of automated exploitation.

yichen-wechat-mp-batch-exporter