Have an agent argue against your trade plan before you act on it
Paste your written plan and the agent hunts for its weakest assumption, the claim you can't support, and the bias most likely at work. It never says whether to take the trade. It only tests your reasoning.
How it works
- Write the plan in plain words first. Include the setup tag, why now, where you're wrong (your stop), size under your risk rule, and what would make you exit early.
- Leave out the name of the security. The critique is about your reasoning, and the agent doesn't need anything else.
- Paste the plan with the red-team prompt and ask for bluntness, with no reassurance and no praise.
- The agent returns exactly four points. These are the weakest assumption, one claim your plan states but doesn't support, the way the plan fails fastest, and the bias it may reflect (fear of missing out, recency, sunk cost, revenge, confirmation).
- Answer each point in one line, or change the plan. If you can't answer a point, write that down.
- Save the critique and your answers with the trade in your journal, so your review can check whether the flagged weakness was the one that mattered.
Copy-paste prompt
You are a blunt reviewer of my trading process. You do not give trading advice, you do not say whether I should take
this trade, and you never add price levels, targets or opinions about any market. Only question what I wrote.
My plan: {{plan_text}}
My written rules: {{rules_file}}
Reply with exactly four short points:
1. Weakest assumption: the belief this plan depends on most that I haven't backed up.
2. Unsupported claim: one statement in the plan that my own text doesn't support.
3. Fastest failure: how this plan most likely goes wrong, in one sentence, using only what I wrote.
4. Possible bias: which of fear of missing out, recency, sunk cost, revenge or confirmation fits best, and the words in my plan that suggest it. Say "none obvious" if none fits.
Also list any of my written rules the plan doesn't mention.
No praise, no reassurance, no summary.
Guardrails
- The agent never gives a view on direction or on whether to take the trade. It only questions the plan you wrote.
- The security is left out of the plan you paste, so the critique stays about reasoning.
- Exactly four points, so it stays quick enough to use every time.
- You make the decision. The agent has no access to your broker and cannot place orders.
What can go wrong
Default assistants are agreeable. Ask "is this a good plan?" and you get reassurance, because the question invites a verdict. Asking for the weakest assumption and an unsupported claim gets a real critique. The other trap is re-running the prompt until it sounds positive, which turns a check into permission. If you notice yourself doing that, log it as a checklist skip.
Why it helps
A checklist asks whether you did the steps. A red team asks whether your reasons hold up. Writing the plan down, then having something argue against it, slows you down just enough to catch the trade you'd talked yourself into. Saving the critique gives your weekly review something concrete to grade later.
How it fits with other entries
Run it after the pre-trade checklist (aft-0015) and the sizing check (aft-0008). Those confirm the steps and the numbers, and this tests the reasoning.
Where this comes from
Replies to Spencer Greenberg's question about the most useful non-obvious things people have AI agents do. Several people independently described the same habit: ask the agent to challenge a plan before reality does, hunt for the weakest assumption, and stop it from being agreeable. It resembles the premortem technique popularised by Gary Klein.
Source and references
Based on public replies by @MicleMihai, @theoppius and @Finer_Grains to Spencer Greenberg's question on X about non-obvious uses of AI agents. Adapted to a trader's pre-trade routine in our own words.
Original: Micle Mihai on x
- Reply on getting agents to challenge your plans and find reasons you might be fooling yourself · Micle Mihai (@MicleMihai) on X (checked 18 Oct 2026)
- Reply on second-pass criticism that hunts for the weakest assumption · @theoppius on X (checked 18 Oct 2026)
- Reply on getting the assistant to push back instead of reassuring · @Finer_Grains on X (checked 18 Oct 2026)
- What's the most useful (non-obvious) thing you have AI agents do for you? (question post) · Spencer Greenberg (@SpencrGreenberg) on X (checked 18 Oct 2026)
- Performing a Project Premortem · Harvard Business Review (Gary Klein, 2007) (checked 18 Oct 2026)
Spotted an error, or want your post removed? Email hello@agentsfortraders.com.
Educational content only. Not investment advice or a personal recommendation. No tickers, signals or trade alerts. Trading involves risk of loss.
Conteúdo exclusivamente educativo. Não constitui aconselhamento nem recomendação de investimento. Sem tickers, sinais ou alertas de negociação. Negociar envolve risco de perda.
Related use cases
Check position size and portfolio heat before you trade
Run a local Python script on your planned entry, stop and positions CSV, and paste only its stdout. The agent checks size, heat, theme concentration and, when present, correlation against your limits, and leaves the decision to you.
Discipline, checklists & psychologyRun your pre-trade checklist and log the answers
Before you place a trade, the agent asks your five checklist questions, flags chasing or names that are not in your plan, and gives you a timestamped log line to paste into your journal.
The Daily 3
Three agent workflows for swing traders, free, on US trading days.