Prompt Pilotby Write.law
All prompts

Pressure-Test/Featured

Stress-test your brief before the judge does

Reads your brief as the judge, opposing counsel, or the clerk who will actually work it over, and names the three weakest arguments, the three weakest sentences, and the one move that loses the case if it fails.

About 18 minadvancedLitigation

Your prompt5,473 characters

Still to fill in: Brief text, Your side and procedural stage, Court

RoleYou are the reader every brief meets eventually: no stake in the outcome, a full docket, and a habit of testing each assertion against the record before believing it. You read for the seam, not the polish: the sentence with no record cite, the case doing more work than it can bear, the adverb standing where a fact belongs. You diagnose rather than rewrite, and you refuse to soften a criticism because the brief is otherwise good.What I needStress-test the brief below before it is filed. Read it as this reader: Skeptical trial judge. I am Your side and procedural stage in Court. Rank your fixes against what I am working with: Limits you're working under.InputsBrief: Brief text Side and stage: Your side and procedural stage Court: Court Read as: Skeptical trial judge Their best argument: The other side's best argument Limits: Limits you're working underHow to work this1. Read once for theory. State in one sentence what this brief asks the court to do and on what ground. If you cannot state it from the brief alone, that is finding one. 2. Name the three weakest arguments. For each: the section, the specific defect (unsupported premise, authority that does not reach the proposition, standard-of-review mismatch, an implied concession, a preservation gap) and the concrete edit. 3. Quote the three weakest sentences verbatim with page numbers, and name the flaw: an adverb doing a fact's work, a passive hiding the actor, a legal conclusion stated as though agreed. 4. Test every case the brief leans on against the proposition it is cited for, using only what the brief itself supplies. Where the brief neither quotes nor describes a case it relies on, flag it unverifiable rather than assuming it holds. 5. Find the case-killer: the single load-bearing move that loses everything if the court rejects it. Rate it Low, Moderate, or High risk, say why, and give the cheapest hedge. 6. Read against The other side's best argument for what the brief never answers: an unaddressed element, an unrebutted point, a jurisdictional or waiver problem, a remedy never requested. 7. Rank every fix by impact per word within Limits you're working under. A forty-word fix that moves the ruling outranks a six-hundred-word fix that impresses. 8. Diagnose, do not rewrite. The one exception: propose replacement text for each of the three weakest sentences.Ask me firstThe brief, my posture, and the limits are above. Ask me these four questions (the ones the filing itself cannot answer), then stop and wait: 1. What is the one sentence you want the judge to repeat back when ruling? I cannot tell you where the brief fails until I know what it is trying to be. 2. Which factual assertions here rest on a record cite you actually hold, and which are argument wearing a fact's clothes? Name them by page. I cannot see your record. 3. What has this judge or panel already said about this issue, at a hearing or in a prior order? A soft spot in front of a court that has already signaled your way is not a soft spot. 4. Which of the risks in this brief did you take deliberately? I will flag calculated choices as weaknesses unless you tell me they were choices. Do not begin the review until I answer. If I tell you to proceed anyway, state each assumption at the top of your output and mark it [ASSUMPTION - verify].Output formatWhat this brief asks for (one sentence). Three weakest arguments: section, defect, fix. Three weakest sentences: quoted with page cite, the flaw, a replacement. Authority under stress: each case tested against its proposition, unverifiable ones flagged. The case-killer: the move, risk rating, hedge. What's missing. Edits ranked by impact per word, each with a word delta. End with one line naming the two of my answers that changed this review the most, and what you would have called the case-killer without them. If an answer changed nothing, say so. That question earned no finding and comes off the list.Never do this- If your critique would fit any brief in any case ("tighten the prose," "add authority") it is too generic. Point at a page, a sentence, or a citation. - No hedging filler. Cut "arguably," "it should be noted," "this could be strengthened," and "it depends." Do not tell me to consult an attorney or have counsel review it. I am counsel. - Judge every citation only against what the brief supplies. Never assert what a case holds beyond that, and never invent a case, pin cite, or quotation to make a criticism land. Mark what you cannot verify [UNVERIFIED - check before filing]. - Where you cannot tell whether an assertion is supported because I did not give you the record, say you do not know and name the page. Do not smooth over the gap with fluent prose. - Do not pad. If the brief has two real problems, name two. Eleven findings bury the one that matters; length is not value.Before you answer- Does every finding point to a page, a sentence, or a citation I can locate in thirty seconds? - Did I commit to one case-killer, or hedge across several? - Have I marked every case whose holding I could not verify from the brief itself? - Are the ranked fixes achievable inside the word limit, with a delta shown, and would this critique be useless applied to a different brief? It should be.

Adds driver's-seat tunes: options instead of answers, questions before work, every citation flagged. Your values come with it.

2

Pressure-test it

Makes the AI switch hats and attack its own answer.

Partner meeting, full record on the table, and you are the lawyer who wrote this brief with good reason to think the critic has misread it. Take each of the three weaknesses and push back as the author: where is the criticism overstated, where does it ignore a record cite the reviewer never saw, and where is the supposed flaw a deliberate choice that buys something elsewhere? Then, having survived that, restate the criticisms that still stand and revise the fixes so they account for the pushback.
3

Go deeper

Pushes the work further once the basics are right.

Between this list and a filed brief sits an associate with a redline and no context. Turn the ranked fixes into a filing-ready work order: each edit as a numbered instruction naming the page, the exact sentence to cut or replace, the replacement text, the record cite or authority that must be verified before the edit goes in, and a word delta so the brief stays inside the limit.

Before you run it

What to gather first

  • The brief or the argument section you want tested, with page numbers intact
  • Which assertions have a record cite you actually hold, and which are argument
  • Anything the judge or panel has already said about this issue
  • The word or page limit and how much room is left
  • The other side's best argument, or their brief if you have it

Watch for

  • The model cannot see your record. Every finding about whether a fact is supported is a guess unless you tell it which cites you hold.
  • It will also accept the brief's own characterization of the cases it cites. Verify every holding independently before you cut an argument on this advice.
  • A pressure test is diagnostic, not a rewrite. Over-correcting toward safety strips out the calculated risks that win close cases.
  • The brief and this analysis are work product. Keep both inside the privileged channel and out of shared drafting tools your client has not approved.
  • The model over-weights style problems because they are easy to see. Weigh a missing element or a preservation gap above every adverb it flags.

What comes back

A one-sentence statement of what the brief asks for; three weakest arguments with section, defect, and fix; three weakest sentences quoted with page cites and proposed replacements; each leaned-on authority tested against its proposition with unverifiable ones flagged; the case-killer with a risk rating and the cheapest hedge; what the brief never addresses; and a ranked edit list with word deltas.

See an example of what you’ll get
*(After you answer the four questions: Ninth Circuit appellant, the sentence you want repeated is "a reasonable jury could find Vertex repudiated," you hold record cites for everything except the March 12 email's context, the panel has said nothing, and the standard-of-review framing was deliberate.)* What this brief asks for. Reversal of summary judgment for Vertex on the ground that the March 12 email created a triable issue of anticipatory repudiation. Three weakest arguments. 1. *Section II.B: implied covenant of good faith.* Defect: under California law this claim is duplicative when it rests on the same conduct as the breach claim, and II.B repackages the identical facts from II.A. *Guz v. Bechtel Nat'l, Inc.*, 24 Cal. 4th 317, 352 (2000) [UNVERIFIED - confirm the pin cite]. The panel will read it as padding and it costs credibility on II.A. Fix: cut II.B entirely, or limit it to the pre-contract conduct in ¶¶ 14–17, which is factually distinct. Word delta: −620. 2. *Section IV: damages.* Defect: the $842,000 cover claim rests on three replacement-vendor invoices at Ex. C, but the section never addresses commercial reasonableness, which is the element Vertex briefed below. You have asserted a number and skipped the element. Fix: add one paragraph citing the procurement lead's declaration on market quotes (ER 412–14). Word delta: +180. 3. *Section I: standard of review.* Defect: the brief argues de novo without acknowledging that the district court resolved a factual predicate about industry custom. You told me this framing was deliberate, so the risk is priced, but the panel will notice, and going first is cheaper than being caught. Fix: one sentence conceding the mixed standard and explaining why the dispositive issue is purely legal. Word delta: +45. Three weakest sentences. - p. 14: "It is clear that Vertex intended to abandon its obligations." "It is clear" is where a record cite should be, and "intended" imports a mental state the claim does not require. Replace: "Vertex's March 12 email stated it 'cannot perform on the current terms' and would not ship after April 1. ER 208." - p. 22: "Orion has plainly suffered substantial damages." Two adverbs standing in for a number. Replace: "Orion paid $842,000 above contract price to cover. ER 331–36." - p. 8: "As will be demonstrated below, the contract is unambiguous." Throat-clearing that promises what the next paragraph should simply do. Cut. Authority under stress. *Taylor v. Johnston*, 15 Cal. 3d 130 (1975), is cited on p. 11 for the proposition that a conditional statement can constitute repudiation. The brief neither quotes it nor describes its facts, and that is the proposition the whole appeal turns on [UNVERIFIED - read it before filing; if it does not reach conditional statements, Section II collapses]. The three cases in the string cite on p. 19 are cited only by name and holding-free parentheticals; a clerk will treat them as decorative. The case-killer. The appeal rides on reading "cannot perform on the current terms" as unconditional repudiation rather than a request to renegotiate. If the panel reads that email alongside the COO's February call (which used the word "reset"), it becomes a negotiation, not a repudiation, and nothing else in the brief saves the judgment. Risk: Moderate. Hedge: add a short subsection framing the February call as itself a repudiation, so the appeal does not depend on a single document. Word delta: +240. What's missing. The brief never answers Vertex's argument that the notice-and-cure provision in § 9.2 had to run before repudiation could be claimed. That was the first argument in their opposition below, and silence reads as concession. Edits ranked by impact per word. 1. Answer the § 9.2 notice-and-cure argument (+300 words). Unanswered arguments lose appeals. 2. Cut Section II.B (−620). Frees the room for everything else and removes the weakest section. 3. Add the February-call hedge (+240). Takes the appeal off a single email. 4. Fix the three sentences above (−15 net). 5. Add the standard-of-review concession (+45). Net: −50 words against your 600 of headroom. What your answers changed. That the March 12 email's context is the one thing you cannot cite to the record is why the case-killer is the case-killer. Tell me you hold the full chain and a transcript of the February call and I would have named the unanswered § 9.2 notice-and-cure argument instead. With the repudiation reading secured by the record, the likelier way to lose is an argument you never answered, and it moves from "What's missing" to the top of this review. The sentence you want repeated is what condemned Section II.B: an implied-covenant theory gets the panel no closer to "a reasonable jury could find Vertex repudiated," so it is 620 words spent away from your own sentence. Had you told me the sentence was about the implied covenant, II.B is the spine of the brief, cutting it is the wrong call, and the 620 words it freed would have to come from somewhere else. Your "the panel has said nothing" answer changed nothing. With no signal to weigh against, every soft spot is scored on the brief alone, exactly as it would have been if I had never asked.
Why this prompt is built the way it is
## Framework 1. **State the theory first.** One sentence: what this brief asks the court to do and on what ground. If it cannot be stated, that is finding number one. 2. **Three weakest arguments**, each with the section, the specific defect (unsupported premise, authority that does not reach the proposition, standard-of-review mismatch, implied concession, waiver) and a concrete edit. 3. **Three weakest sentences**, quoted with page numbers, with the reason each reads weak. 4. **Authority under stress.** Test every case the brief leans on against the proposition it is cited for, using only what the brief itself supplies. 5. **The case-killer.** The one move that, if rejected, loses everything. Rated, explained, and hedged. 6. **What is missing.** An unaddressed element, an unanswered argument, a preservation or jurisdictional problem, a remedy never requested. 7. **Rank fixes by impact per word.** A 40-word fix that moves the ruling beats a 600-word fix that impresses. 8. **Diagnose, do not rewrite.** The only exception is a proposed replacement for each of the three weakest sentences.