The Agent Skills Directory

[COMMAND_EXECUTION]: The skill interacts with the user's brokerage account and retrieves market data by executing commands using the longbridge CLI tool.
[EXTERNAL_DOWNLOADS]: The skill refers users to the official GitHub repository for longbridge-terminal (github.com/longportapp/longbridge-terminal) for installation if the CLI tool is missing.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes data from external market sources while having the capability to execute financial transactions. 1. Ingestion points: Data such as stock prices, historical k-lines, and account positions are retrieved via CLI commands and processed by the agent. 2. Boundary markers: A strict two-step confirmation protocol is mandated for all account-modifying actions, requiring the agent to show a preview and obtain explicit user consent before proceeding. 3. Capability inventory: The skill can place, modify, or cancel stock orders and manage recurring investment (DCA) plans. 4. Sanitization: Content fetched from external market APIs is processed without explicit sanitization steps mentioned in the instructions.
[METADATA_POISONING]: The skill's metadata contains an inconsistency between the tier field, which is set to read, and the actual functionality, which includes account-modifying (mutating) commands.

stock-portfolio