Prompting Like a Lawyer — Specification, Worked Examples and the Output Contract
Briefing, taught as lawyers already brief juniors: a specification, an output contract that forces honesty, and a ceiling diagnostic.
- 90 min
- 60 taught + 30 lab
- Direct the machine, decode it
- Build-along
- Free vs. paid comparison
- Your laptop · free tier
The lab · 30 min · Specification + output contract
The Internship Note Prompt. The internship-note prompt demonstrated end to end, its uncertainty flags read as a work list.
The full lab planThe hook
“'Summarise this judgment' asks for luck. A specification with an output contract is a brief a junior could execute, and unlike the junior, the machine never comes back to ask what you meant.”
What you'll be able to do
- Write a specification for a legal task (task, audience and register, context, constraints, format) that says everything that matters once, and nothing else.
- Attach an output contract that forces the model to show its uncertainty: an authorities table with a 'checked?' column, mandatory flags on anything unverified, and standing permission to say 'I am not certain'.
- Steer form with worked examples, two or three exemplars where the output must match a house style, and know when examples help, when they constrain, and where the vendors disagree.
- Handle refusals and hedges professionally, iterate by changing one thing at a time, know when to abandon a chat and start clean, and diagnose whether a disappointing answer is your prompt's ceiling or the model's.
On the syllabus
- The specification: task, audience, context, constraints, format, once
- The output contract: authorities table, uncertainty flags, permission to abstain
- Worked examples: two or three exemplars steer form
- Refusals, hedges and iteration: reframe, argue both sides, restart derailed chats
- The folklore autopsy, and the diagnostic: model's ceiling or prompt's?
The session
How to write a prompt that works: state the task, the reader, the facts, the constraints and the format, and add an output contract that makes the model list its authorities as unchecked and flag what it is unsure of. The session also tests popular prompt tricks against the evidence and compares a free model with a paid one on the same task.
Why it matters
A reusable template for the term: a specification plus an output contract that turns any research or drafting ask into checkable work with its uncertainty on the surface. And a habit that saves money and blame: before you curse the model or buy a better one, fix the brief.
What you can do on Monday
Take the next real ask you get (a moot issue, a project topic, an internship request) and run it through your saved specification-plus-contract template instead of typing a one-liner. File the output with its 'checked?' column as your work list.
What you leave with
The skill
Brief the machine like a junior: a specification that says everything once, an output contract that forces the uncertainty into the open, and exemplars where the form must match.
The insight
Prompt quality comes from how clearly you had already thought about the task, never from magic words. The contract matters more than the phrasing, because it makes the model's ignorance visible instead of fluent.
The moment you'll remember
The build-along's third round. The one-line ask produced confident mush and the specification produced structure; then the output contract lands and the same model confesses, with a table reading 'checked?: no' against every authority and a flagged 'I am not certain' where it had been smoothest. The room audibly reacts to the confidence coming off. Students arrive believing prompting coaxes better answers and leave having watched it force honest ones.
In this session
- 01
The specification, field by field: the task as a deliverable; the audience and register ('for a supervising advocate who has not read the file'), which is the real use of a 'role', since personas change tone and focus rather than accuracy; the context the model cannot know; the constraints (jurisdiction, date, length, exclusions); the format. Then the 2026 discipline: say each thing once and stop, because every vendor warns that bloated, repetitive, shouty prompts degrade output. The colleague test: if a colleague with no context would be confused, so will the model.
- 02
The output contract is the lawyer's move. Require an authorities table (authority | proposition | paragraph | checked?) with 'checked?' set to no throughout, because the model cannot check. Require anything uncertain to be flagged rather than smoothed over. Grant standing permission to say 'I am not certain', which vendors report sharply reduces invented detail; Session 2 derives the mechanism, which is that the line changes the scoring the model was trained to optimise. The contract cannot make the model honest. It makes the uncertainty visible, and turns fluent output into a work list.
- 03
Worked examples: for anything that must match a shape (a clause in the firm's style, a case comment in a journal's format, an issue list in your supervisor's style), two or three exemplars beat any description. The 2026 caveat: Anthropic recommends a handful of diverse examples, Google says always include them, OpenAI says reasoning models often do better without, and examples that conflict with your instructions mislead. The rule that survives is that exemplars steer form and never supply accuracy. If the output slavishly copies the wrong thing, the examples are why.
- 04
Refusals and hedges: 'I can't provide legal advice' usually means the ask sounded like a client matter; reframe it as an academic fact pattern and it proceeds. The deeper skill is hedge management. An 'it depends' answer is often the correct legal answer wearing a lazy face, so make it earn its keep: demand the best view, the strongest contrary view, and the facts that would flip the answer. And do not bully. Current guidance is explicit that aggressive, capitalised, threatening prompt language degrades compliance.
- 05
Deliberate iteration: change one variable per run and log what changed, or you have learned nothing from the improvement. Know when to stop iterating. Models get lost in long corrective conversations and rarely recover; one 2025 study measured an average 39% drop from a clean single ask across 15 models. After two failed corrections, the professional move is a fresh chat with everything you have learned folded into a better first prompt.
- 06
The folklore autopsy: a tip, a threat and 'please' have no reliable aggregate effect (Wharton Prompting Science Reports, 2025), and 'think step by step' produces marginal-to-no accuracy gain on models that already reason internally, at real cost in time. Today's free tiers all run thinking models, so the projector demo of that folklore shows nothing. What replaced it is structure you can check: IRAC with paragraph pinpoints into supplied text, rather than a narration of the model's hidden reasoning, which research shows is not a faithful transcript.
- 07
The diagnostic that runs through the course: when output disappoints, is it the model's ceiling or your prompt's? The order of operations is fixed: specification first, context next (Session 3), the harness after that (Session 4), only then the model (Session 5). Today's comparison shows why: the same well-specified reshaping task, run on the free tier and on a paid frontier model, comes back equivalent. Almost everything students blame on the model is a prompt or context problem, and the course will show the two places where it is the model after all.
- 08
The computational-thinking lens, named: a specification is abstraction, stripping the request to what matters, the same move as extracting a ratio; a staged sequence is decomposition, issue-spotting applied to your own ask. The pillars appear here as working habits rather than a syllabus; they are the reason a legal education is a head start at this craft.
The build-along
The room builds it and keeps it.
The class builds the thing with the presenter, the mechanism is explained as it is built, and a named skill closes the hour.
The internship note, built in four rounds
Build-alongThe task
The task: a supervising advocate wants a note on a live question by tomorrow. Round one is the one-line ask most of the room would type today; its output goes on screen. Before each further round the room votes on which addition will help most, then dictates it: specification, then context and constraints, then the output contract.
Why it works
Each addition changes the output for a reason the mechanism explains: the specification narrows the distribution; the context replaces the statistically average assumption with your facts; the contract changes what the model optimises toward, from smooth confidence to visible uncertainty.
Built live
Four runs, side by side on the projector, one addition at a time. The fourth output carries an authorities table with 'checked?: no' down the column, the first honest draft of the day.
The skill
Build the brief before you type it: task, audience, context, constraints, format, then the contract that forces the doubt into the open.
The folklore autopsy
DemonstrationThe room predicts
The room votes on which of four prompt 'boosters' will improve a fixed answer: a ₹500 tip, a threat, 'please', and 'think step by step'. Most rooms bet on at least two.
What is going on
The 2025 evidence: no reliable aggregate effect from tips, threats or politeness; marginal-to-nothing from CoT incantations on models that already reason before answering, which is what every 2026 free tier runs.
Shown live
The same question runs plain, tipped, threatened and step-by-stepped, live. The answers are equivalent, and that null result, predicted on screen before the run, is the finding.
The skill
Spend your effort on the specification and the contract rather than on incantations. When a technique claims magic, ask for its evidence and its date.
Making 'it depends' earn its keep
DemonstrationThe room predicts
A genuinely contestable question goes up, and the room drafts the follow-up they would send when the model hedges.
What is going on
Hedging is trained-in safety, but it yields to structure: a contract demanding the best view, the strongest contrary view and the facts that would flip the answer converts a shrug into legal analysis, without bullying, which current models punish.
Shown live
The hedge appears on cue, the structured follow-up runs, and the answer comes back as two argued positions with conditions: the shape a supervisor wants.
The skill
Never accept mush and never bully. Demand both sides and the switching conditions; that is what 'it depends' owes you.
Free versus paid · you watch this one
The clause rewrite — where paying changes nothing
The task, on both: Rewrite a dense indemnity clause in plain English for a client, to a tight specification with two exemplar rewrites supplied: a pure reshaping task with a full brief.
The presenter runs the same prompt on a free tier and on a stronger model or higher reasoning-effort tier, side by side. You watch; nobody needs a paid plan.
Free tier
The free tier, given the full specification and exemplars, returns a clean, accurate, correctly-registered rewrite. Nothing material to fix.
Stronger model / higher effort
The presenter's paid frontier model, given the identical prompt, returns an equivalent rewrite, differently worded and no better. The room compares them blind and cannot reliably pick the paid one.
A well-specified reshaping task saturates the free tier. When output on a task like this disappoints, the ceiling you have hit is your prompt's; fix the brief before you reach for a wallet. Sessions 3 and 5 show the two places where the ceiling is the model's.
The legal thread
The output contract is your first professional artefact: a table saying 'checked?: no' against every authority is the honest starting state of every AI-assisted document, and converting those entries to 'yes' by hand is what candour to the tribunal demands before your name goes on anything (Session 6 drills the check; Session 8 the certification). Notice what the built prompt never contained: a client's name. Abstraction before prompting starts here and becomes doctrine in Sessions 4 and 7.
The lab · 30 minutes · demonstrated in class, repeated on your own
The Internship Note Prompt
The internship-note prompt demonstrated end to end, its uncertainty flags read as a work list. Afterwards, students build the same prompt around their own task and submit it.
1 · The demonstration
In class, on one realistic task.
On
Gemini or ChatGPT (free tier)
The exact prompt
You are drafting for a supervising advocate who has not read the file. Task: a research note on whether a licensor can forfeit the full security deposit when a licensee exits a leave-and-licence agreement before the lock-in period. Context: Maharashtra; commercial premises; no negotiated liquidated-damages clause. Constraints: Indian law only; note any point where the position is unsettled; do not invent authority. Format: (1) issue in one sentence; (2) short answer in three; (3) analysis under headings; (4) a table — authority | proposition | paragraph | checked? — with 'checked?' as 'no' for every row, because you cannot verify; (5) a section titled 'What I am not certain of', which must not be empty. If you are not certain a case exists, say so instead of naming one.
Point at
Point at three things when it returns: the table's 'checked?: no' column (the honest starting state of every AI draft); the populated 'not certain' section (permission to abstain, working); and any authority in the table, which nobody can trust yet; the 'checked?: no' column says so, and Session 2 shows, live, where authorities like these come from.
What comes back
A well-structured note with two to four authorities of mixed reality in the table and a useful uncertainty section. The substance will be plausible and unverified. Say so, and leave it unverified: verification is Session 6's drill, and the flags are today's point.
If it misbehaves
The pre-captured note from the same prompt (fallback folder, S1). If the live model refuses the framing as legal advice, prepend 'For a law-school exercise:' and show the room the reframing move; it is part of the course content.
2 · Afterwards, on your own
The same technique on a task of your own; the deliverable goes to the platform.
Afterwards, on your own: pick a real task (a moot issue, a seminar topic, an internship-style ask), build the specification field by field, attach the contract verbatim, run it on a free tier, and save the prompt itself to a prompt bank. The saved prompt plus its flagged output is the platform submission.
Step by step
- 01Choose a real task of your own: the moot issue you are arguing, the seminar topic you are writing, or the kind of note your internship asked for.
- 02Draft the specification field by field: task as a deliverable, audience and register, context the model cannot know, constraints (jurisdiction, date, length, exclusions), format.
- 03Attach the output contract: the authorities table (authority | proposition | paragraph | checked?), mandatory uncertainty flags, and standing permission to say 'I am not certain'.
- 04Run it on a free tier and read the output against the contract: did every promised section appear, and what did the contract force the model to admit?
- 05Save the prompt itself to your prompt bank, not only the output, and mark in the margin every claim you would have to verify before real use.
Tools · free tier
- Gemini or ChatGPT (free tier)
- The shared prompt-bank template (posted before the session)
- The specification + output-contract skeleton from the build-along
Free tier
One long prompt and one or two refinement runs, comfortably inside every free tier, text-only. If rate-limited, draft the full prompt offline in your notes app and run it once when the window resets; the drafting is the assessed skill.
If the free tier locks you out
Draft the full prompt offline in your notes app and run it once when the window resets. The drafting is the assessed skill; the run is confirmation.
3 · The reveal
Before the contract round runs, the room votes on which addition will improve the output most. The reveal is the fourth output itself: an authorities table with 'checked?: no' in every row and a populated 'what I am not certain of' section, compared line by line with round one.
Deliverable
A saved, reusable specification-plus-contract prompt for a real task, its first output with every unverified claim flagged, and one line on what changed after the first run. The first lab deliverable submitted on the course platform.
Feeds the Lab deliverablesRun of show · 30 minutes
- 0–20 min — The demonstration: the internship-note prompt built in four rounds, the room voting before each addition, the outputs compared side by side.
- 20–26 min — The contract read closely: what it forced the model to admit, and which claims a lawyer would verify first.
- 26–30 min — The assignment: the same build on a task of your own, due on the platform before Session 2; questions.
Verification checklist
- The specification states task, audience, context, constraints and format, each stated once.
- The output contract includes the authorities table with 'checked?' defaulting to no, and an uncertainty section that must not be empty.
- The output was read against the contract and deviations noted.
- The prompt is saved in reusable form, and every unverified claim in the output is flagged.
For the instructor · before the session
Why this shape
Craft is learned by building, so the course opens with its first build-along. The room constructs the internship-note prompt with the presenter, voting on each addition before it runs and watching the output change, and leaves the first evening holding a reusable template. An experiment here would only prove students cannot prompt yet, which nobody needs proven.
- Pre-run the four build-along stages that morning and screenshot each output; the improvement arc is the lesson, and models drift.
- Load the fallback folder: the four staged outputs, the folklore-autopsy runs, and the pre-captured demo note.
- Have the paid-tier account signed in for Compare A, with the identical clause-rewrite prompt staged in both tiers.
- Launch the Session 1 poll deck with the class size; run the warm-up question early; the build-along votes and the folklore predictions run on phones.
- Post the shared prompt-bank template (a plain document with fields) to the class channel so nobody builds theirs from a blank page.
- Post a starter-task list (moot-style issues and internship-style asks) to the class channel; students without a live task of their own borrow one for the practice segment.
Key sources & cases
OpenAI, GPT-5-era prompting and reasoning guidance (developers.openai.com, read 2026-08-27)
The vendor's positions, cited to the live pages: prompting reasoning models to 'think step by step' is unnecessary; try prompts without examples first on reasoning models; leaner prompts outperformed (vendor-internal figures; quote as vendor claims); contradictory instructions damage the newest models most. Re-check each cohort; the pages moved hosts in 2026 and will move again.
Anthropic, prompting best practices and 'reduce hallucinations' documentation (platform.claude.com, read 2026-08-27)
Worked examples (3–5, diverse) for format and tone; permission to say 'I don't know' as a first-line hallucination reduction; 'dial back' aggressive prompt language on current models; quotes-first grounding for long documents. Vendor guidance rather than measured findings, and labelled as such in class.
Meincke, Mollick, Mollick & Shapiro, Prompting Science Reports 1–3 (Wharton, 2025)
The folklore autopsy's evidence base: politeness effects are contingent and unpredictable (Report 1); chain-of-thought prompting yields marginal-to-no gains on reasoning models at 20–80% time cost (Report 2); tipping and threatening have no significant aggregate effect (Report 3). Verified to the arXiv abstracts 2026-08-27.
Zheng et al., 'Personas in System Prompts Do Not Improve Performances of LLMs' (EMNLP Findings 2024)
162 personas, four model families, 2,410 factual questions: no accuracy improvement. The verified basis for teaching 'role' as audience-and-register rather than as an accuracy lever.
Laban et al., 'LLMs Get Lost in Multi-Turn Conversation' (2025)
An average 39% drop from single-turn to multi-turn performance across 15 models; models that take a wrong turn 'do not recover'. The evidence behind 'after two corrections, start a fresh chat with a better first prompt'. Verified to the abstract 2026-08-27.
Chen et al. (Anthropic), 'Reasoning Models Don't Always Say What They Think' (2025)
Visible chains of thought verbalised the cue the model used in often under 20% of cases. This is why the course teaches 'demand a checkable argument with pinpoints' instead of 'read the model's mind'.
Jeannette Wing, 'Computational Thinking', CACM 49(3) (2006)
Decomposition and abstraction as general habits of mind: the lens, retained from the course's first design, for why legal training transfers to this craft.
Readings
- Anthropic, 'Prompting best practices' — skim the sections on examples, roles and reducing hallucinations (link in the class channel; read the current page, not a cached one)
- OpenAI, 'Reasoning best practices' — the two paragraphs on step-by-step prompting and few-shot examples
- Wharton Prompting Science Reports 1–3 (2025) — abstracts only: the evidence that buried the folklore
- Laban et al., 'LLMs Get Lost in Multi-Turn Conversation' (2025) — abstract
- Your own last five real prompts, re-read against today's template; bring the worst one to Session 3
Sixteen hours, one professional discipline.
Craft, competence and verification, practised until they are habits you could defend in court.