The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads several components from the okx organization on GitHub, including the onchainos CLI installer, the meteora-plugin binary, and management scripts (launcher.sh, update-checker.py). These downloads are performed from vendor-controlled infrastructure.
[REMOTE_CODE_EXECUTION]: The setup process in SKILL.md fetches an installation script (install.sh) from the okx/onchainos-skills repository and executes it. This process includes a security check that verifies the script's SHA256 hash against a downloaded checksum file before execution to prevent tampering.
[COMMAND_EXECUTION]: The skill makes extensive use of the system shell and the onchainos CLI to manage wallet state, fetch balances, and broadcast transactions to the Solana network. These operations are within the expected scope of a blockchain-focused AI agent skill.
[DATA_EXFILTRATION]: While the skill interacts with external APIs (dlmm.datapi.meteora.ag) and Solana RPC nodes (api.mainnet-beta.solana.com, rpc.ankr.com), these communications are limited to protocol-specific data and do not involve sensitive local files or credentials. Financial transactions are only executed after user approval using the --confirm flag.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external data such as pool names, token symbols, and on-chain account data from the Meteora API and Solana blockchain. A data trust boundary is explicitly defined in the documentation, instructing the agent to treat this content as untrusted and purely informational to mitigate potential injection risks from third-party protocol metadata.

meteora-plugin