skill-04-nft-management

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.75). The required workflow ingests outsider-authored free text when it resolves and validates NFT metadata by fetching off-chain content from IPFS/Arweave (e.g., JSON fields like name/description/attributes) and when it uses marketplace aggregation via Reservoir/OpenSea/Blur APIs, which can include arbitrary text from third-party listings/metadata that the LLM reads into context.

MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).

• Direct money access detected (high risk: 1.00). The skill explicitly describes blockchain transaction actions and wallet-like capabilities: minting tokens ("Execute mint; verify token received"), buying/selling on marketplaces, private transaction submission, setting marketplace approvals, and creating/signing from Token-Bound Accounts (TBA) which "can hold ETH, ERC-20" and "Sign transactions from TBA via ERC-1271". These are direct crypto/blockchain financial operations (wallet signing, sending transactions, managing on-chain assets), so it grants direct financial execution authority.