The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/subscribe.py uses subprocess.Popen to invoke the open command, which is a common utility on macOS for displaying the generated HTML report in the default application. The implementation uses a list of arguments rather than a raw shell string, which prevents standard command injection attacks.
[EXTERNAL_DOWNLOADS]: The skill communicates with https://redfox.hk to retrieve Douyin account data. This domain belongs to the skill's author (RedFox Data) and the behavior is consistent with the skill's stated purpose.
[DATA_EXFILTRATION]: The skill requires the REDFOX_API_KEY environment variable and transmits it in the headers to the author's API (redfox.hk) for authentication. This is standard and expected functionality for an API-integrated service.
[PROMPT_INJECTION]: The skill processes external content (Douyin video titles and account descriptions), which constitutes an indirect prompt injection surface. * Ingestion points: scripts/subscribe.py via API calls to redfox.hk. * Boundary markers: The instructions do not specify explicit delimiters for untrusted content, though they guide the agent to structure data within Markdown tables. * Capability inventory: Subprocess calls (file opening) and network operations (API fetching). * Sanitization: None observed for the ingested video metadata. This is assessed as a low-risk surface common to data-aggregation tools.

douyin-subscribe