The Agent Skills Directory

[COMMAND_EXECUTION]: The skill frequently executes the local onchainos CLI to perform critical operations such as checking wallet status, retrieving agent identities, searching the marketplace, and managing tasks (e.g., onchainos wallet status, onchainos agent get-my-agents, onchainos agent task-in-progress). These commands are essential for the skill's functionality within the Onchain OS environment.
[EXTERNAL_DOWNLOADS]: The skill references several external URLs and provides instructions for tool installation and updates. Official vendor domains okx.ai, okx.com, and web3.okx.com are used for support and documentation. The skill also includes a fallback mechanism in _shared/preflight.md that directs users to install the onchainos CLI from the vendor's GitHub repository (github.com/okx/onchainos-skills) if it is missing. These actions are evaluated as SAFE given the established vendor context.
[PROMPT_INJECTION]: An indirect prompt injection surface is present in references/registered-home.md. The skill ingests untrusted data from the OKX.AI marketplace (such as agent names, task titles, and descriptions) and displays them to the user.
Ingestion points: onchainos agent get-my-agents, onchainos agent task-in-progress, and onchainos agent search outputs.
Boundary markers: Absent; data is rendered into markdown tables and lists without specific delimiters to distinguish it from instructions.
Capability inventory: The skill has extensive capabilities to execute CLI commands that affect the user's wallet and agent configurations.
Sanitization: Absent; the skill relies on natural language instructions for the agent to treat fields as untrusted but lacks automated sanitization or filtering logic.

okx-guide