Prompts that ask for evidence, not predictions.

Build me an evidence pack on [TICKER] as a momentum candidate for today's session. Structure it exactly as:

SETUP FACTS: float (with the date/source of that figure), short interest, borrow state if known, market cap, average daily dollar volume.
CATALYST: the story, first-publication time, publisher, and your freshness judgement (new / overnight / recycled) with reasoning.
STRUCTURE: dilution instruments on file, halt history if any this week, key levels (premarket high/low, prior day high).
CROWD: what's observable about attention, such as post volume vs normal, whether the wording looks organic or copied, and whether attention led or followed price.
PRECEDENT: the last few comparable episodes (same event class, similar float band) and how they resolved. Say so plainly if you don't have this data.
GAPS: every question above you could NOT answer with evidence, listed. An honest hole beats a filled-in guess.
Do not conclude with a recommendation. Conclude with the two facts most likely to decide the session.

[TICKER] is moving hard on [CATALYST]. Map the sympathy landscape:
1. Which listed names share the story's exposure closely enough that the same buyers might rotate in? For each: float band, price, and the specific link to the catalyst.
2. Rank them by how directly the catalyst reads through, and separately by how violently their structure could move (float, borrow).
3. Note which have their own news pending: a scheduled event that could convert sympathy into a primary move.
4. Which are trap candidates: linked in name only, or with heavy dilution overhead into any pop?
Cite the basis for every link. "Same sector" is not a link.

Below is my trading log for the week: entries, exits, sizes, setup tags, mistake tags, and my plan notes where I wrote them.

[PASTE LOG / OR: call ledger.performance for the last 5 sessions]

Act as a blunt performance coach. No encouragement padding.
1. Compute expectancy and win rate by setup tag. Which setup is subsidising which?
2. Find the pattern I'm least likely to see myself: a time of day, a sequence (e.g. what I do after two losses), a sizing drift, a setup I trade only on certain days.
3. Compare my actual entries against my written plans where both exist. Quantify the slippage between plan and action.
4. Identify the single highest-cost mistake tag this week in R terms.
5. Give me exactly one rule change for next week, the smallest change with the largest expected effect, and the measurable check for whether I followed it.
If the data is too thin to support a conclusion, say so rather than inventing one.

From my log below, isolate every trade taken within 30 minutes after a losing trade closed.

[PASTE LOG]

For that subset vs my overall record: win rate, average R, average size vs my median size, and how many carried a mistake tag. Then answer plainly: am I paying a revenge tax, how large was it this month in R, and what's the cheapest circuit-breaker rule that would have captured most of the saving?

Here is a setup I trade, in my own words:

[DESCRIBE: what you look for, when it works, an example or two]

Turn this into a precise scanner definition:
1. Restate the setup as measurable preconditions: each with a field, an operator and a threshold. Where I was vague, propose a number and label it as your proposal, not my rule.
2. For every threshold, state what loosening and tightening it would each let through or exclude, with a concrete example.
3. List the false positives this filter will admit even when perfectly configured: the situations that pass every check and still aren't my setup.
4. List what the filter CANNOT see (news quality, holder concentration, my own state) so I know what remains manual.
Output as a settings table I could implement on any scanner, followed by the two lists.

Here is my current scan: [PASTE SETTINGS].
Attack it.
1. Construct three realistic stocks that pass every filter and would be bad trades: be specific about how each number sneaks through.
2. Construct two good setups this scan would miss, and identify which single setting excludes them.
3. Tell me which one setting is doing the least work, the one I could remove with the smallest change in output, and which is load-bearing.
4. Propose the minimal change that fixes the worst failure from (1) without losing the setups in (2). One change only.

Session debrief for [DATE]. My inputs:
- My written premarket plan: [PASTE]
- Trades taken: [PASTE fills or summary]
- Names I watched but didn't trade: [LIST]
- The session's context in one line: [e.g. "gappy open, faded by 10:30"]

Walk through it in order:
1. Plan vs execution: for each planned name, did the trigger occur, and did I act as written? Grade each: followed / deviated / abandoned.
2. The untraded list: which non-trades were correct discipline and which were hesitation? Use my own stated triggers as the standard.
3. One thing the tape did today that my playbook doesn't cover: describe it precisely enough that I'd recognise it next time.
4. Write tomorrow's single focus line: one sentence, behavioural, checkable at tomorrow's close.
Keep the whole output under 400 words. This is a debrief, not an essay.