The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill includes instructions to install numerous Node.js packages from the @metaplex-foundation scope and facilitates the download of project templates through the mplx toolbox template command.
[COMMAND_EXECUTION]: The documentation relies on the mplx CLI to perform on-chain operations such as registering agents, creating collections, and executing token launches.
[DYNAMIC_EXECUTION]: The skill features capabilities for executing raw, base64-encoded blockchain instructions via the mplx toolbox raw CLI command and the Core execute instruction in the SDK, which are intended for flexible smart contract interaction.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data sources, creating a potential surface for indirect prompt injection when ingesting untrusted metadata or RPC responses.
Ingestion points: Off-chain metadata JSON files (e.g., metadata.json) and blockchain asset data retrieved via DAS API methods like getAsset.
Boundary markers: The skill utilizes structured JSON formats for data exchange, although specific delimiters for preventing instruction leakage during prompt construction are not detailed in the examples.
Capability inventory: The environment supports shell command execution, network communication with RPC and Irys storage providers, and reading local configuration files (e.g., id.json).
Sanitization: While the skill promotes the use of standard schemas, it does not explicitly document validation or sanitization logic for data retrieved from external URLs or RPC endpoints.

metaplex