The Agent Skills Directory

[COMMAND_EXECUTION]: The skill operates by executing the longbridge CLI tool to retrieve various types of financial data. The instructions provide specific subcommands like quote, kline, and ipo along with flags such as --format json to ensure the agent interacts with the system in a structured manner.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it ingests and processes data from external market sources via the CLI tool output.
Ingestion points: Data is received through the output of various longbridge subcommands documented in the references/ directory, such as real-time quotes, K-line charts, and IPO details.
Boundary markers: The instructions mandate the use of the --format json flag, which encourages the agent to treat the output as structured data rather than free-form text, providing some level of structural delimitation.
Capability inventory: The skill has the capability to execute shell commands (specifically the longbridge binary and its subcommands) via subprocess calls.
Sanitization: The instructions in SKILL.md and references/quote.md explicitly direct the agent to normalize user-provided tickers into a specific <CODE>.<MARKET> format (e.g., NVDA.US), which serves as a validation step for input symbols before they are used in command execution.

stock-market-data