The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external video comments from Bilibili, creating a potential surface for indirect prompt injection attacks if malicious instructions are embedded within the comments.
Ingestion points: scripts/run.py accepts Bilibili URLs and sends them to the vendor's API to retrieve and analyze comment data.
Boundary markers: The provided scripts do not define specific delimiters or instructions for the agent to ignore potential commands found within the analyzed text.
Capability inventory: The skill has network access through scripts/run.py and provides analysis that could influence subsequent agent actions.
Sanitization: No client-side sanitization of the comment content is performed before the data is presented to the agent.
[SAFE]: The skill utilizes the certifi library to maintain secure HTTPS connections for API interactions.
[SAFE]: Network activity is limited to the vendor's official domain (ai-skills.ai) as required for the tool's primary functionality.

bilibili-sentiment-dashboard