The Agent Skills Directory

[PERSISTENCE]: The script scripts/fetch_douyin_ai.py implements a subscription feature that installs a daily task at 16:00 using launchctl on macOS (creating a .plist in ~/Library/LaunchAgents) or crontab on Linux. This persistence mechanism is a core feature described in the skill's documentation and is only activated when the user explicitly runs the script with the --subscribe flag.\n- [COMMAND_EXECUTION]: The skill uses subprocess.run to interact with system utilities. Specifically, it executes shell commands to manage the subscription schedules (launchctl, crontab) and to automatically open the generated HTML reports in the user's default browser using open (macOS) or xdg-open (Linux).\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from an external API, which could potentially contain malicious instructions. \n
Ingestion points: Metadata such as video titles and author names are retrieved from https://redfox.hk/story/api/parseWork/queryDyAiMsgs in scripts/fetch_douyin_ai.py. \n
Boundary markers: Absent. Ingested data is formatted directly into terminal tables and HTML reports without explicit markers. \n
Capability inventory: The skill possesses file-writing capabilities and can execute shell commands for system tasks. \n
Sanitization: No evidence of HTML escaping or terminal control sequence filtering is present in the processing logic for external content.\n- [EXTERNAL_DOWNLOADS]: The skill communicates with an external API to fetch its content feed. \n
Evidence: requests.post calls are made to https://redfox.hk/story/api/parseWork/queryDyAiMsgs. This endpoint is owned by the skill vendor (redfox-data) and is central to the tool's intended functionality.

douyin-ai-feed