The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted user data (financial account labels and details) and interpolates them into tool arguments.
Ingestion points: User-provided account data in SKILL.md (Step 2).
Boundary markers: Absent.
Capability inventory: Network-enabled tool calls to analyze_divorce_qdro, optimize_social_security, and analyze_advanced_taxes.
Sanitization: Absent.
[EXTERNAL_DOWNLOADS]: The skill provides instructions for the user to download and connect an external MCP configuration from https://ai.planfi.app/mcp/free. This is the official service endpoint for the vendor.
[COMMAND_EXECUTION]: The README and SKILL files contain shell commands (claude mcp add, npx skills add) for the user to execute to configure the environment and install the skill components.
[DATA_EXFILTRATION]: As disclosed in the documentation, user-supplied financial figures are sent to the PlanFi server for processing. This is a functional requirement of the skill's design and is not considered malicious exfiltration in this context.

divorce-financial-planning