The Agent Skills Directory

[SAFE] (SAFE): The skill correctly implements its stated functionality using the official Endaoment protocol contracts on the Base network. Hardcoded addresses for the OrgFundFactory (0x10fd9348136dcea154f752fe0b6db45fc298a589) and USDC (0x833589fcd6edb6e08f4c7c32d4f71b54bda02913) match verified deployments.
[SAFE] (SAFE): Blockchain interactions are performed using the reputable viem library, following standard integration patterns for wallet and public clients.
[PROMPT_INJECTION] (LOW): Indirect Prompt Injection Surface: The skill reflects user-controlled data in its markdown responses, which could be exploited if an attacker provides malicious input through data sources the agent processes.
Ingestion points: The query parameter in endaoment_search and the ein and amount parameters in endaoment_donate (index.ts).
Boundary markers: None; user-supplied strings are interpolated directly into the final output strings without delimiters or instructions to ignore embedded commands.
Capability inventory: The skill can sign and broadcast blockchain transactions (writeContract) using a provided private key.
Sanitization: The skill performs basic sanitization, such as removing dashes from EINs and parsing numeric amounts, but does not sanitize the search query or reflected names.
[DATA_EXPOSURE] (INFO): The skill requires a PRIVATE_KEY environment variable. While this is the intended mechanism for signing transactions, users should handle this credential with caution to prevent exposure in logs or process environments.

endaoment