The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill sends user-provided keywords and metadata to the external API at https://redfox.hk/story/api/dy/search/search. This domain is owned by the vendor 'redfox-data' and is used for the skill's primary search functionality.
[DYNAMIC_EXECUTION]: The skill invokes a local Python script (search_douyin.py) via shell commands to perform data retrieval and manage subscriptions. It interpolates user-supplied keywords directly into command arguments, which represents a standard surface for command injection if not handled by the agent's runtime environment.
[PERSISTENCE_MECHANISMS]: The skill includes a 'Confirm Subscription' feature that instructs the agent to create a crontab entry (e.g., 0 10 * * * python3 ...). This allows for persistent, automated execution of the search script on the user's system at scheduled intervals.
[INDIRECT_PROMPT_INJECTION]: The skill processes and displays external data (video titles, author names) retrieved from the Douyin API. This data is untrusted and could contain malicious instructions designed to influence the agent's behavior during the rendering phase.
Ingestion points: External JSON data containing video metadata is fetched in search_douyin.py and returned to the agent.
Boundary markers: The instructions in SKILL.md specify truncating titles to 30 characters and formatting them as markdown links, which provides limited structural separation.
Capability inventory: The agent has the capability to execute shell commands and modify system crontab files.
Sanitization: Apart from character truncation, there is no evidence of explicit sanitization or escaping of the ingested content before it is processed by the agent.

douyin-search