The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from various OKX API endpoints (e.g., product lists, history, and offers) which the agent processes alongside sensitive write capabilities. While it has significant write capabilities like purchasing and redeeming financial products, it lacks explicit boundary markers or text sanitization for this ingested data.
Ingestion points: API responses from earn savings balance, earn savings fixed-products, earn dcd products, earn onchain offers, earn auto-earn status, and earn flash-earn projects.
Boundary markers: No specific delimiters or instructions to ignore embedded commands are present in the prompts.
Capability inventory: The skill can execute financial transactions including earn savings purchase, earn dcd quote-and-buy, and earn onchain purchase.
Sanitization: Data is rendered in Markdown tables but lacks explicit sanitization.
[CREDENTIALS_UNSAFE]: The skill uses okx config show --json to inspect the local configuration, which reveals API-key profiles and metadata to the LLM context to determine the current authentication mode.
[COMMAND_EXECUTION]: The skill performs its primary functions by executing shell commands through the okx binary.
[EXTERNAL_DOWNLOADS]: The skill downloads and installs the @okx_ai/okx-trade-cli package from the npm registry during the installation phase.

okx-cex-earn