Skip to Content
AgentQL SkillPrompt Patterns

Prompt Patterns

Use these patterns when you want the host to stay inside the current public AgentQL contract without over-explaining transport details.

Market read prompt

Use the AgentQL skill to return the latest price, source status, and freshness for this ethereum token.

Market context prompt

Use the AgentQL skill to resolve the market first, then summarize the market profile and the latest price.

Monitoring prompt

Use the AgentQL skill to monitor this market with live delivery and summarize only meaningful price changes.

MCP-style prompt

Use the AgentQL skill through MCP-style tool routing. Start with one market tool first and only widen when the first response is insufficient.

Native RPC prompt

Use the AgentQL skill to call the hosted EVM RPC compatibility route for eth_chainId and return only the result.

Prompt rules

  • Name the user job before the transport.
  • Prefer market-read language for market-read tasks.
  • Prefer native JSON-RPC language only for chain data tasks.
  • Keep endpoint truth externalized through agentql.tomo.inc/nodes.
  • Avoid legacy BSC-first or partner-only framing in new prompts.
Last updated on