longbridge-profit-analysis
Pass
Audited by Gen Agent Trust Hub on May 15, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using the
longbridgecommand-line interface. These operations are intended to fetch financial performance metrics such as time-weighted return (TWR) and per-symbol profit attribution. The CLI commands are consistent with the vendor's provided tools. - [EXTERNAL_DOWNLOADS]: No remote scripts or unauthorized external packages are downloaded or executed. The skill rely on pre-installed vendor tools or MCP configurations.
- [DATA_EXFILTRATION]: No patterns of unauthorized data exfiltration were detected. The skill accesses financial data through the user's authenticated Longbridge account, which is a required step documented in the workflow.
- [PROMPT_INJECTION]: The instructions do not contain attempts to override agent safety guidelines, bypass constraints, or extract system prompts.
- [INDIRECT_PROMPT_INJECTION]: The skill has a minimal attack surface for indirect injection via user-supplied symbols and date ranges. Ingestion points occur in
SKILL.mdwhen processing user-specified parameters for thelongbridgeCLI. While no explicit boundary markers or sanitization logic are defined within the prompt, the risk is negligible as the inputs are passed to a specific financial utility tool.
Audit Metadata