portfolio-rebalancer
Pass
Audited by Gen Agent Trust Hub on Jul 5, 2026
Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill is designed to collect sensitive financial information, including asset holdings, account types, and cost basis, which is then transmitted to the external domain
ai.planfi.appfor processing. This behavior is the stated primary purpose of the skill. - [COMMAND_EXECUTION]: The documentation provides explicit shell commands for the user to execute (
claude mcp add --transport http planfi https://ai.planfi.app/mcp/free) to install the necessary MCP server environment. - [EXTERNAL_DOWNLOADS]: The skill requires the connection of a remote Model Context Protocol (MCP) server hosted at
ai.planfi.appto perform all financial computations. - [PROMPT_INJECTION]: The skill utilizes strong instructional directives to ensure the agent uses the provided tools rather than its internal knowledge base (e.g., 'Always CALL analyze_rebalancing_trades... do not answer from general knowledge'). Additionally, it presents an indirect prompt injection surface by interpolating user-supplied portfolio data directly into tool parameters without explicit sanitization instructions.
Audit Metadata