The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes subprocess.run in scripts/subscribe.py to orchestrate the internal report generation process by calling scripts/generate_report.py. Additionally, scripts/generate_report.py uses platform-specific commands (such as open, start, or xdg-open) to automatically display the resulting HTML report in the user's web browser. It also attempts to read shell profile files (e.g., .zshrc, .bashrc) to retrieve the required REDFOX_API_KEY environment variable for the user's convenience.
[EXTERNAL_DOWNLOADS]: The generated HTML reports include a reference to the html2canvas library from the jsDelivr CDN (cdn.jsdelivr.net). This is a well-known and trusted service used to enable the report's "Export as Image" functionality.
[INDIRECT_PROMPT_INJECTION]: The skill processes and summarizes external data from TikTok, such as video descriptions and summaries. This creates an indirect prompt injection surface. The risk is mitigated by the skill's use of structured output formats (Markdown tables and headers) and explicit labeling of the AI-generated summaries as inferences based on public data.
Ingestion points: External TikTok metadata is retrieved via API calls to redfox.hk in scripts/subscribe.py.
Boundary markers: The agent uses Markdown tables and clear subheaders to delineate external content from instructions.
Capability inventory: The skill performs network operations via Python's socket and ssl modules and writes local files in the output/ directory.
Sanitization: Content is summarized and escaped for rendering in both Markdown and HTML formats.

tiktok-subscribe