The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it fetches and processes untrusted content (video titles and profile descriptions) from the Douyin platform. If this content contains malicious instructions, the agent might inadvertently execute them while generating its 'Content Insight' report.
Ingestion points: External data is ingested through the JSON output of the fetch_user_works.py script.
Boundary markers: Absent. The agent is not instructed to use delimiters or ignore potential commands within the fetched video titles.
Capability inventory: The agent can execute local Python scripts and perform network operations to the vendor's API.
Sanitization: No validation or sanitization is performed on the data returned by the API before it is analyzed by the agent.
[COMMAND_EXECUTION]: The skill relies on the execution of a local Python script (scripts/fetch_user_works.py) to perform its primary function. This script is implemented using standard libraries and employs argparse to safely handle the user-supplied account identifier, reducing the risk of local command injection.
[DATA_EXFILTRATION]: The skill performs network requests to doubaoya.com and transmits a user-provided API key (DOUBAOYA_API_KEY) for authentication. While this involves sending sensitive credentials over the network, it is the intended and necessary behavior for the skill's stated purpose of interacting with the DoubaoYa service. No unauthorized access to local sensitive files was detected.

douyin-works-crawler