The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/cultural_tourism_douyin_report.py script uses subprocess.run with shell=True to install Linux crontab entries. The command string is constructed using the --keyword argument without sufficient validation or escaping, which could allow for arbitrary shell command execution if a specially crafted keyword (e.g., containing ; or &&) is provided.
[PERSISTENCE]: The skill implements a 'daily subscription' feature that creates persistent background tasks. On macOS, it installs a LaunchAgent (~/Library/LaunchAgents/com.qoder.cultural-tourism-douyin-feed.plist), and on Linux, it modifies the user's crontab. The macOS plist file explicitly stores the REDFOX_API_KEY in its EnvironmentVariables section, making the secret persistent on the file system.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data (titles, author names) from an external API (redfox.hk) and renders it into an HTML report. The Python script uses simple string .replace() calls to interpolate this data into the HTML template without proper HTML escaping or sanitization. This creates an attack surface where malicious content from the API could inject scripts (XSS) or influence the agent's interpretation of the report content.
Ingestion points: Data is fetched from https://redfox.hk/story/api/parseWork/queryDyPlayletMsgs in scripts/cultural_tourism_douyin_report.py.
Boundary markers: None used in the HTML report generation.
Capability inventory: The script can write files to the filesystem and execute shell commands to manage system services.
Sanitization: The sanitize_title function only addresses Markdown table characters ([ and |) and does not provide proper HTML entity encoding.

cultural-tourism-douyin-feed