The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied parameters including currency symbols, network names, and fiat amounts to construct requests for the OpenFinance API.
Ingestion points: Parameters are extracted from user prompts such as amount, chain, currency, and baseCurrency in SKILL.md.
Boundary markers: The skill relies on structured JSON inputs for the MCP tool get_onramp_url rather than unstructured text interpolation.
Capability inventory: The skill identifies available chains (Ethereum, Polygon, Base, etc.) and performs smart routing to external providers. It does not possess file-writing or arbitrary command execution capabilities.
Sanitization: The skill contains explicit validation logic (e.g., the coin/network matrix for Onramp.money) to ensure only supported combinations are processed.
[DATA_EXFILTRATION]: No sensitive data access was detected. The skill generates URLs for external payment providers (Moonpay and Onramp.money) where the user completes KYC and payment in their own browser, keeping sensitive financial data outside the agent's context.
[COMMAND_EXECUTION]: The skill does not perform any local shell command execution or system-level operations. It operates entirely through REST API interactions with the OpenFinance backend.

openfin-onramp