The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill heavily processes untrusted external data retrieved via longbridge news, longbridge filing, and various WebSearch queries for financial metrics (e.g., TAM, industry reports). This data is used for sentiment scoring and financial analysis, creating a surface for indirect prompt injection where malicious content in a news article or corporate filing could attempt to influence the agent's behavior.
Ingestion points: The skill ingests text from news articles, corporate filings, and web-based industry research reports.
Boundary markers: The instructions do not define specific delimiters or warnings for the agent to ignore instructions embedded within the processed financial content.
Capability inventory: The skill has the capability to execute shell commands via the longbridge CLI and perform web searches, though its metadata restricts it to a read_only risk level.
Sanitization: No specific sanitization or content validation logic is described for the incoming text streams before they are analyzed by the NLP components.

longbridge-intel