Skip to main content
Agent Skills describe intent-level actions using deterministic JSON rather than Stellar SDK calls, CLI commands, RPC methods, or XDR. Each specification contains:
  • Stable slug, semantic version, protocol and category.
  • Explicit stellar:testnet and/or stellar:pubnet support.
  • Authentication mode and safety requirements.
  • Named actions with JSON input and output schemas.
  • Machine-readable error codes, descriptions and retryability.
  • Catalog, specification and MCP endpoints.
Initial skills cover Stellar payments, Soroban invocation, Aquarius swaps, and scheduled disbursements.

​
Discover

Filters are query, protocol, network, and action. Unknown slugs return SKILL_NOT_FOUND.

​
Agent behavior

  1. Search by user intent.
  2. Select a skill supporting the required network.
  3. Validate proposed input against the named action schema.
  4. Respect simulationRequired, walletAuthorizationRequired, and networkMustBeExplicit.
  5. Request a wallet signature only after presenting the simulated effect.
  6. Branch on the returned error code and retryable value; never parse prose to decide whether to retry.
The console Skills tab provides filters, readable safety/action/error views, and a copyable raw specification for integration testing.