The Agent Skills Directory

[Indirect Prompt Injection] (HIGH): The skill processes untrusted external content (bid URLs and names) retrieved from the blockchain via RPC calls. This data is then interpolated into transaction templates for the 'Bankr' tool.
Ingestion points: Data returned from eth_call via curl (specifically URLs and names in bid history) and direct user input.
Boundary markers: None. Untrusted strings are interpolated directly into prompts for the Bankr tool.
Capability inventory: Financial transaction execution (via Bankr), arbitrary curl commands to RPC endpoints.
Sanitization: No evidence of sanitization or validation of the URLs or names retrieved from the blockchain.
[Command Execution] (LOW): The skill uses curl and jq to interact with the public Base RPC endpoint (https://mainnet.base.org). This is a legitimate requirement for the skill's functionality but should be monitored for unexpected endpoint changes.
[Financial Risk] (INFO): The skill facilitates the spending of USDC. Users should verify contract addresses (0x7309...A176) independently before authorizing large approvals.

qrcoin