Write the trading preferences that shape every conversation and background run.

System Prompt

Your system prompt is the brief you give your agent before it starts working. It shapes how Cabal reasons, what it prioritizes, and what kind of output it produces. Every conversation and background run uses the active agent's system prompt automatically.

Where to edit

Open Agents from the dashboard, choose the agent, then write your system prompt on its dedicated tab. It is freeform text -- write it however makes sense to you.

What to include

Think about what you'd tell a human analyst on their first day:

Risk tolerance -- How aggressive or conservative should Cabal be?
Position sizing -- Max percentage of portfolio per trade? Dollar limits?
Focus areas -- DeFi infrastructure? Memecoins? Specific sectors?
Entry/exit rules -- When to take profit? Where to cut losses?
What to avoid -- Tokens, sectors, or strategies you don't want

Good example

I'm moderately aggressive. Max 5% of portfolio per trade. Focus on DeFi infrastructure and memecoins only if 3+ alpha channels are calling it. Take profit at 50%, stop loss at 20%. Don't touch anything with less than $500K daily volume. When in doubt, suggest rather than execute.

This works because it is specific. Cabal can apply these rules consistently across conversations and background runs.

Bad example

Make me money. Be smart about trading. Don't lose too much.

This doesn't work because there is nothing actionable. Cabal can't turn "be smart" into a trading strategy. It needs concrete preferences to reason against.

Research-focused example

I track Solana DeFi and on-chain activity. I care about whale wallet movements, new protocol launches, and liquidity shifts. When I ask about a token, always include volume trends, holder distribution, and any relevant on-chain signals. Don't suggest trades unless I specifically ask -- I use Cabal primarily for research.

This works even if you never enable execution. Your system prompt shapes all of Cabal's output, not just trade proposals.

System prompt vs. guardrails

Your system prompt and guardrails do different jobs:

	System Prompt	Guardrails
What they do	Shape reasoning and output	Enforce hard limits on execution
Where they live	Agent page	Agent page
Enforcement	Guidance -- Cabal follows them but they are not hard stops	Hard stops -- Cabal cannot execute beyond them
Example	"Max 5% per trade" as a preference	Max position size as an enforced limit

If you want Cabal to prefer small positions, put it in your system prompt. If you want Cabal to be physically unable to take large positions, set it as a guardrail.

Tips

Be specific. Numbers and rules work better than vibes.
Update them. Your system prompt should evolve as your strategy changes.
Test them. Ask Cabal a question and see if the response reflects your preferences. If not, tighten the language.
Don't duplicate guardrails. If you've already set a max position size as a guardrail, you don't need to repeat it in the system prompt.

FAQ

How long should my system prompt be?

As long as it needs to be. A few sentences covering risk, sizing, and focus areas is enough to start. Add detail as you learn what Cabal gets right and wrong.

Does the system prompt affect background runs?

Yes. Every run -- whether triggered by Chat, a watch, or a routine -- uses the selected agent's system prompt.

Can I have different system prompts for different situations?

Yes, by using separate agents. Each agent has its own system prompt, model defaults, control profile, workspace, and operating history.