AI Trading Prompt Pack: 14 Prompts for Analysis, Discipline and Review

Most “AI trading” prompts ask the model to predict. That is the wrong task, and it produces the wrong kind of confidence. These prompts ask the model to describe, structure, check and challenge — work it can do, with outputs a human can verify.

Use them as starting points. Adapt the instrument, timeframe and rules to your own process. Do not paste account numbers, balances, broker credentials or personal data into a consumer AI tool.

The prompt is not the system. The system is the workflow: a human writes the plan, the model checks it, and the human owns the decision.

How To Use This Pack

  1. Start with the observations-only prompt. Do not ask for a bias yet.
  2. Check every level and number against the chart or the broker platform.
  3. Use the scenario and challenge prompts to improve the plan, not to choose the trade.
  4. Use the review prompts on anonymized journal entries.
  5. Keep the final decision, the order and the accountability with the trader.

A prompt is only as good as the input and the verification. If you cannot check the output in a few minutes, do not use it for a live decision.

1. Observations Only

You are a chart-reading assistant. Do not predict. Using only this image, list:

1. Instrument, timeframe and timestamp (or “not visible”). 2. Swing structure: rising, falling or range. 3. Position in the visible range: upper, middle or lower third. 4. The five nearest visible levels, with their labels. 5. The indicator values printed on the chart. 6. Bar character: spread, volume, close location, overlap, tails, follow-through.

If a value is not visible, write “not visible.” Do not infer. Do not add levels that are not on the chart.

Check: every level and number against the chart. Delete anything you cannot see.

2. Four-Pass Chart Review

Apply the four-pass review to this chart:

Pass 1 — Clean price: swing structure, range position, break acceptance or rejection. Pass 2 — Character and volume: spread, volume, close location, overlap, tails, follow-through. Pass 3 — Context: higher-timeframe position versus its mean and range; agreement or conflict. Pass 4 — Conditionals: bullish, bearish and no-trade branches, each with a trigger, invalidation and first target.

Separate observation from interpretation. Label any statement that is an interpretation. Do not invent levels.

Check: the four passes, the labels, the invalidation.

3. Three Scenarios With Invalidation

Using only the observations above, write three scenarios: primary, secondary and no-trade. For each, give:

– the trigger, – the entry, – the stop, – the first target, – the invalidation, – the evidence that would change your mind.

Do not assign probabilities unless I ask. If a scenario has no invalidation, say so explicitly.

Check: every scenario has an invalidation; the no-trade branch is real.

4. Challenge My Bias

Here is my current read and plan:

[paste the plan]

Argue the opposite case. Identify the three assumptions the plan depends on and the evidence that would falsify each one. Do not rewrite the plan. List the weaknesses.

Check: the counterarguments; whether any assumption is untested.

5. Multi-Timeframe Reconciliation

Compare the higher-timeframe and lower-timeframe reads below. Where do they agree, where do they conflict, and which one should govern a trade taken on the lower timeframe? State the conflict explicitly. Do not average it away.

Higher timeframe: [paste observations] Lower timeframe: [paste observations]

Check: the conflict is stated, not hidden; the governing timeframe is explicit.

6. Plan Omission Check

Review this one-page plan. Flag:

– any scenario without an invalidation, – any stop that sits inside the last swing, – any target nearer than the stop, – any risk number that exceeds the daily limit, – any rule that is not testable, – any level that is not traceable to a source.

Return a list of required changes, not a rewrite.

Check: each flagged item against the plan and the chart.

7. Position Size Arithmetic Check

Here are the inputs:

– Instrument and point value: [ ] – Stop distance: [ ] – Account equity: [ ] – Risk percentage: [ ] – Position cap: [ ] – Number of accounts: [ ]

Calculate the maximum contracts for each account. Show the arithmetic. Flag any account where the result exceeds the cap or the daily loss headroom. Do not recommend a trade.

Check: the inputs; the point value; the arithmetic. Recalculate independently.

8. Order Review

Here is my plan and my working order list:

[paste plan] [paste orders]

For each order, tell me:

– whether it matches a scenario in the plan, – whether it has a stop, – whether the stop is beyond the structure, – whether the target is at least 2R, – whether the size is within the cap, – what the total position and risk would be if every order filled.

Return a list of required changes: keep, reduce, tighten, cancel, or flatten on a condition.

Check: every order in the broker platform; the aggregate; the stops.

9. Pre-Mortem For Revenge Trading

I have just taken a full loss. My rules are:

[paste rules]

Ask me the five questions I must answer before I take another trade. Do not comfort me and do not tell me to trade. If the rules say stop, say so.

Check: the rules; the daily loss headroom; whether the next trade is on the plan.

10. Journal Pattern Analysis

Here are 20 anonymized journal entries. Identify:

– the three most frequent deviations from the stated plan, – the emotional states that precede them, – the times of day when they occur, – the setups with the lowest compliance.

Quote the entries that support each pattern. Do not give advice; show the evidence.

Check: the quotes; the sample size; whether the pattern is causal or coincidental.

11. Weekly Review

Here is my week of anonymized journal entries. Produce:

1. Aggregate metrics: trades, win rate, average R, profit factor, largest win, largest loss. 2. Average compliance and the number of trades below 8/10. 3. The most common rule broken, with quotes. 4. One hypothesis for what to change next week, with a way to measure it.

Do not give generic advice. Use only the entries.

Check: the metrics; the quotes; whether the hypothesis is testable.

12. Backtest Critique

Here is my backtest design:

[paste hypothesis, rule, data, sample, costs, results]

Critique it against these questions:

– Is there look-ahead bias? – Is the sample large enough? – Are costs and slippage realistic? – Is there survivorship bias? – How many variations were tested? – Is there an out-of-sample or walk-forward test? – Does the result survive a 2× cost assumption? – Is the drawdown acceptable?

Do not tell me the strategy is good. Tell me what would falsify it.

Check: the data, the sample, the cost assumptions, the out-of-sample test.

13. Vendor Diligence

I am evaluating an “AI trading” tool or signal vendor. Here are its claims:

[paste claims]

Build a diligence checklist covering: the exact strategy, live versus backtested track record, independent audit, costs and fees, maximum drawdown, execution and custody, data quality, regulatory status and exit terms.

Flag any claim that cannot be verified from the information provided.

Check: the vendor’s evidence; the regulatory status; the terms.

14. Research Summary Critique

Summarize this study in five bullet points: sample, period, instruments, method, and result after costs. Then list what the summary omits and what would be needed to replicate the result.

[paste abstract or paper]

Check: the sample, the costs, the method, the replication.

The Verification Checklist

Before you act on any AI output, check:

  • [ ] Every level is visible on the chart or traceable to a named source.
  • [ ] Every number has been recalculated independently.
  • [ ] Every scenario has an invalidation.
  • [ ] The no-trade branch is present.
  • [ ] The risk is within the per-trade and daily limits.
  • [ ] The position size is within the cap.
  • [ ] Every order has a stop.
  • [ ] No account numbers, balances or credentials were shared.
  • [ ] A named human approved the plan and the order.

If any box is unchecked, the output is a draft, not a decision.

What Not To Ask

  • “What should I trade?”
  • “Will price go up or down?”
  • “What is the best entry?”
  • “Give me a signal.”
  • “How much can I make?”
  • “Is this guaranteed?”
  • “Place the order for me.”

These questions invite the model to do something it cannot do responsibly. The answer may be fluent, confident and wrong.

Disclaimer

This prompt pack is educational. It is not financial, investment, legal or tax advice, and it is not a recommendation or solicitation. Trading involves substantial risk of loss. AI output can be inaccurate or incomplete; verify every level, figure and assumption on the chart and in the broker platform. Past performance and backtests do not guarantee future results.

Start With a Quick Brief

Book a 30-minute strategy call. We'll diagnose your challenge and map the right AI-powered skill set for your goals.

Book the Call