The Agent Skills Directory

[SAFE]: The skill follows a non-custodial security model by generating transactions in an unsigned hex format. This design ensures that the agent never has access to user private keys, requiring the user to sign and submit transactions via an independent operator skill.
[SAFE]: The scripts include built-in safety logic, such as warning the user if slippage is set above 10% and using parseFloat to validate numeric inputs for amounts and slippage.
[PROMPT_INJECTION]: The skill exposes a surface for indirect prompt injection (Category 8) because it processes untrusted user input like token tickers and swap amounts to interact with DEX APIs.
Ingestion points: User parameters are passed as command-line arguments to scripts/preview-swap.js and scripts/create-swap.js.
Boundary markers: The skill definition in SKILL.md mandates that the agent must always show the user the pool pair, amounts, slippage, and fees, and requires explicit confirmation before calling the transaction creation script.
Capability inventory: The skill uses Node.js scripts to communicate with the SaturnSwap API via a local client library.
Sanitization: Input amounts are sanitized using parseFloat to ensure they are valid numbers before being processed.

saturnswap-market-swap