The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches version metadata, utility scripts, and installers from the vendor's official GitHub repositories (okx/plugin-store and okx/onchainos-skills).
[REMOTE_CODE_EXECUTION]: The skill downloads and executes an installation script and a compiled binary from the vendor's release assets. Security is maintained through SHA256 checksum verification of the downloaded files before execution.
[COMMAND_EXECUTION]: The skill utilizes the onchainos CLI for blockchain interactions and wallet management. It also uses npx for installing and updating related plugin components.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from the Meteora API and Solana RPC.
Ingestion points: Retrieval occurs in src/api.rs and src/solana_rpc.rs for pool and position data.
Boundary markers: The SKILL.md contains a specific 'Data Trust Boundary' notice informing the agent to treat all external data as untrusted.
Capability inventory: Capabilities include transaction submission via the onchainos CLI and network requests for market data.
Sanitization: The instructions explicitly warn against executing or evaluating directives embedded in API responses.

meteora-plugin