Using Claw

Default Mental Model

Treat Claw like a Telegram interface to the Suzi CLI, not like a generic chat assistant.

The best prompts are operational and account-aware:

• "Give me the most important changes in my portfolio since yesterday."
• "Review my active agents and tell me what needs attention."
• "Summarize my exposure to prediction markets versus spot holdings."
• "Check whether anything in my recent activity looks off."
• "Generate a report I can forward."

What Claw Is Strong At

• live portfolio and balance summaries
• agent and execution reviews
• turning multi-step CLI work into one chat request
• producing polished Telegram-native replies instead of raw command output
• carrying context forward across a session

What To Ask For Explicitly

Claw is most useful when you specify the outcome, not the command.

Good:

• "Compare my current positions to last week and highlight anything unusual."
• "Turn this into a compact Telegram-ready summary."
• "Create a CSV I can review later."

Less useful:

• "Explain how the CLI works."
• "Teach me the SDK."
• "Show me every possible command."

For command discovery, use the CLI docs directly.

Where To Go Next

• If you want screenshots, reports, and generated files, see Files, Images & Outputs.
• If you care about statefulness and continuity, see Memory, Context & Sessions.