The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external blockchain data and user-provided inputs, creating a surface for indirect prompt injection.
Ingestion points: Data enters through blockchain state queries and token metadata fetching from scripts/check-balance.ts and scripts/check-balances.ts.
Boundary markers: Safety policies in agents/openai.yaml require explicit user confirmation and safety checks for all operations.
Capability inventory: The skill possesses significant capabilities, including token transfers and smart contract invocations via starknet_transfer and starknet_invoke_contract tools.
Sanitization: The skill relies on standard SDKs (starknet.js, avnu-sdk) and agent-level confirmation prompts for data handling and execution safety.
[EXTERNAL_DOWNLOADS]: The skill interacts with external blockchain services and APIs.
Fetches token metadata and verified token lists from AVNU's official API.
Connects to well-known Starknet infrastructure providers such as Alchemy and Cartridge to interact with the network.
These network operations are necessary for the primary function of the skill and target established ecosystem services.

starknet-wallet