The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted external data retrieved from the Stellar blockchain, such as ledger entries, transaction histories, and contract events. This creates a surface for indirect prompt injection if the agent processes malicious strings embedded in blockchain records.
Ingestion points: Data enters the context via rpc.getEvents, rpc.getLedgerEntries, and Horizon endpoints like server.transactions().forAccount(publicKey). This data originates from on-chain activity which can be controlled by any network participant.
Boundary markers: The skill lacks explicit instructions or delimiters (like XML tags or markdown blocks with 'ignore' directives) to differentiate between verified instructions and untrusted data fetched from the chain.
Capability inventory: The skill possesses the capability to simulate transactions (rpc.simulateTransaction), submit signed transactions to the network (rpc.sendTransaction, server.submitTransaction), and poll for state changes.
Sanitization: There is no evidence of content sanitization or validation of the fetched blockchain data before it is presented to the agent's context.
[EXTERNAL_DOWNLOADS]: The skill relies on official Stellar development resources and standard package registries.
Dependencies: The skill instructions demonstrate usage of the official @stellar/stellar-sdk package.
Network References: The skill connects to official network endpoints including https://soroban-testnet.stellar.org, https://horizon.stellar.org, and public infrastructure repositories on GitHub.

data