The Agent Skills Directory

[PROMPT_INJECTION]: The skill implements a defensive pattern to protect against indirect prompt injection. It wraps untrusted data returned from the OpenSea API (such as NFT names and descriptions) in clear boundary markers (--- BEGIN OPENSEA API RESPONSE ---), instructing the agent to treat such content as data rather than instructions.
[EXTERNAL_DOWNLOADS]: The skill provides a script to fetch an API key from the official OpenSea authentication service and recommends installing the official @opensea/cli package via npm. These operations target the vendor's own verified services and tools.
[CREDENTIALS_UNSAFE]: The skill handles an optional OPENSEA_PRIVATE_KEY for local signing (SIWE). Instructions and scripts explicitly manage these credentials through environment variables or restricted local files (mode 600), following standard security practices for wallet-enabled tools.

opensea-api