The Agent Skills Directory

[PROMPT_INJECTION]: The skill identifies a potential surface for indirect prompt injection as it processes untrusted external data, though it includes several mitigation strategies.
Ingestion points: Untrusted data enters the agent context through longbridge news, longbridge filing, longbridge topic, and a potential WebSearch fallback for breaking news.
Boundary markers: The skill utilizes structured output templates and mandatory classification buckets (e.g., catalyst, regulatory, strategic) to organize external data, which helps maintain context separation.
Capability inventory: The skill executes the longbridge CLI tool for data retrieval and has write capabilities through longbridge topic create and create-reply commands.
Sanitization: It specifically instructs the agent to filter out speculative "hype vocabulary" and provides fact-only takeaway constraints to prevent the echoing of malicious or biased content.
[COMMAND_EXECUTION]: The skill makes routine use of the longbridge CLI tool to interact with financial data and community platforms. These operations are consistent with the vendor's intended functionality and do not involve suspicious shell patterns or privilege escalation.

longbridge-content