The Agent Skills Directory

[DATA_EXFILTRATION]: The script scripts/tiktok_hot_search.py attempts to read sensitive shell configuration files (~/.zshrc, ~/.bashrc, ~/.bash_profile, ~/.profile) to find the REDFOX_API_KEY. While this is intended for user convenience as a fallback for environment variables, accessing these files is generally considered a sensitive operation.
[INDIRECT_PROMPT_INJECTION]: The skill processes and summarizes external data from TikTok (video descriptions, topic names, and user profiles). This creates a surface for indirect prompt injection if an attacker places malicious instructions in TikTok metadata.
Ingestion points: Search results from the RedFox API, specifically fields like content, userName, and topicName processed in scripts/tiktok_hot_search.py.
Boundary markers: Data is rendered into structured Markdown tables; video summaries are truncated to 60 characters to limit context length.
Capability inventory: The skill can perform network requests to redfox.hk, write to the local filesystem (JSON and HTML output), and execute shell commands to open the browser.
Sanitization: The skill uses html.escape during HTML report generation to prevent XSS and applies string truncation for Markdown output.
[COMMAND_EXECUTION]: The script scripts/generate_report.py uses subprocess.run to execute platform-specific commands (open, start, or xdg-open) to automatically open the generated HTML report in the user's default browser.
[EXTERNAL_DOWNLOADS]: The HTML report generated by scripts/generate_report.py includes a script tag that fetches the html2canvas library from the cdn.jsdelivr.net CDN at runtime.

tiktok-hot-search