longbridge-portfolio
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill extensively uses the
longbridgeCLI to retrieve account data and perform market operations. Mutating commands that involve financial risk, such asorder buy/sell/cancelanddca create/update, are protected by a mandatory two-step protocol. The agent is instructed to provide a detailed preview and wait for explicit verbal confirmation from the user before executing these specific commands. - [EXTERNAL_DOWNLOADS]: The skill references the installation of the
longbridge-terminalCLI from the vendor's GitHub repository (github.com/longportapp/longbridge-terminal). It also points to a local installer script (install-cli.sh) within the skill's relative directory structure to assist with initial setup. - [INDIRECT_PROMPT_INJECTION]: The skill functions by ingesting external data from account positions, performance records, and market quotes, which could potentially contain adversarial content.
- Ingestion points: External data enters the agent context through the JSON output of CLI commands like
longbridge positionsandlongbridge kline. - Boundary markers: The instructions enforce the use of strict output templates and industry-standard reporting frameworks (e.g., Brinson attribution, MPT) to structure the processed data.
- Capability inventory: The skill possesses shell execution capabilities through the
longbridgebinary, including the ability to perform trade mutations. - Sanitization: The skill mitigates risks by requiring the agent to process data into structured formats and predefined templates before presenting it to the user.
Audit Metadata