← lib/lib/prompt

Prompt library.

The ones I actually reach for — title and use-case at a glance, full text on preview, copy in one click. Take what you need.

62 / 62 prompts
shelf
PRM-001prove

Prove it, don't describe it

Before accepting any agent claim that work is done — run the receipt loop first.

PRM-002prove

Refute that this is merge-ready

Fresh-context reviewer — prompted to break the change, not approve it.

PRM-003write

Clarify the question before drafting

When an essay or note keeps rewriting itself — the question is probably wrong.

PRM-004plan

Structured literature pass

For papers and concepts — extract the claim, method, and what would falsify it.

PRM-005plan

Scope before implement

When a task is vague — force a minimal plan with explicit non-goals.

PRM-006prove

Review until the round is empty

Meta-prompt for multi-round review loops — stop only when a pass finds nothing.

PRM-007ship

Rebuild the landing page from positioning

The landing page doesn't pass the 10-second test — who, problem, outcome, difference, first value moment.

PRM-008build

Design a frictionless feature

Adding or reshaping a feature — it must feel obvious on first use and reduce cognitive load.

PRM-009audit

Maximize feature value

A feature works but doesn't feel indispensable — audit cognitive, emotional, and monetary value.

PRM-010build

Ship high-leverage improvements autonomously

You want the agent to pick and execute the highest-ROI product work without a task list.

PRM-011plan

Research-backed product concept

Starting a new product — need honest concept doc grounded in real user pain, not category clichés.

PRM-012audit

Activation and time-to-value audit

Signups don't convert to meaningful value fast enough — find what's missing or fragile in the first journey.

PRM-013audit

Elevate from vibe-coded to product-grade

The app feels assembled from startup tropes — diagnose why and prioritize standards upgrades.

PRM-014ship

Premium UI and copy pass

The product needs quiet confidence — calm, precise, desirable without hype or noise.

PRM-015ship

Product-led viral loop

GTM needs organic distribution — design a share mechanism rooted in the product workflow, not referral spam.

PRM-016prove

Pre-mortem before you ship

Before a release, launch, or merge you can't easily undo.

PRM-017prove

Falsify the assumption ledger

When the plan rests on beliefs nobody has tested yet.

PRM-018prove

Audit every user-facing claim

Before shipping copy, UI labels, or marketing that states outcomes.

PRM-041prove

Blameless post-mortem

After an incident, failed launch, or outage — learn without scapegoating.

PRM-042prove

Hypothesis-driven debug

When a bug is elusive — stop guessing and run a scientific isolation loop.

PRM-060prove

SBAR agent session handoff

When resuming a long task, switching agents, or handing off to a fresh context window.

PRM-061prove

ReAct debug loop

When an agent is guessing at fixes — force thought, action, and observation each step.

PRM-062prove

Reflexion after failure

When an approach failed — critique what went wrong before retrying, not immediately rerolling.

PRM-019plan

Jobs-to-be-done interview synthesis

Before building — when you need the real job, not stated preferences.

PRM-020plan

Riskiest-assumption sprint plan

When you have many unknowns and need to pick what to learn first.

PRM-021plan

Competitive teardown

Before positioning or building — map what alternatives actually deliver.

PRM-043plan

Opportunity Solution Tree

When roadmap items aren't tied to outcomes — map bets to measurable results.

PRM-044plan

User story map

When the backlog is a flat list — visualize the journey and slice releases.

PRM-053plan

Mom Test problem interview

Before building — run interviews that surface past behavior, not compliments.

PRM-054plan

Working Backwards PR-FAQ

Before a big build — force customer-visible narrative before specs or wireframes.

PRM-059plan

Plan-and-execute gate

Before an agent writes code — separate planning from execution with an explicit gate.

PRM-022build

Empty, loading, and error states

When a feature ships without designed non-happy paths.

PRM-023build

First-session onboarding slice

When new users don't reach first value in one sitting.

PRM-024build

Accessibility implementation pass

Before calling a UI feature done — keyboard, focus, contrast, labels.

PRM-045build

Form and input design

When a flow is mostly fields and buttons — where most products lose users.

PRM-046build

Microcopy pass

When UI works but labels, buttons, and errors sound like placeholders.

PRM-047build

Notification design

When the product pings users — email, push, in-app — without a clear policy.

PRM-025audit

Nielsen heuristic evaluation

Quick structured UX audit without user tests.

PRM-026audit

HEART metrics diagnosis

When you need to know which metric is broken and why.

PRM-027audit

Kano feature classification

When prioritizing features — must-have vs performance vs delight.

PRM-048audit

Product-market fit survey

When you need one honest signal — would users be disappointed if this disappeared?

PRM-049audit

Cognitive walkthrough

When a specific task flow fails — step through what the user thinks at each action.

PRM-050audit

RICE prioritization

When the backlog is opinion wars — score impact, confidence, effort, and reach.

PRM-028write

Pyramid Principle outline

When a doc or essay feels scattered — lead with answer, then support.

PRM-029write

One-page decision memo

When a decision needs a shareable, async-readable brief.

PRM-030write

Explain to engineer and skeptic

When an idea is clear to you but lands poorly with technical or skeptical readers.

PRM-034write

BLUF the async update

For email, Slack, or status posts where most readers stop after the first line.

PRM-035write

Toulmin argument map

When a persuasive draft feels convincing but the logic chain is invisible.

PRM-036write

Orwell–Zinsser clarity edit

When a draft exists but reads bloated, abstract, or like it was written to impress.

PRM-037write

Plain language rewrite

When a doc is legally correct or technically accurate but ordinary readers won't finish it.

PRM-038write

Reverse outline

When you've written a lot but can't tell if the structure actually works.

PRM-039write

Pinker curse-of-knowledge diagnostic

When experts wrote it and non-experts bounce — you know too much to notice.

PRM-040write

Strip AI cadence, restore human voice

When the draft reads polished, generic, and interchangeable with every other AI-generated text.

PRM-055write

Nut graf and lede

When the opening doesn't earn the read — bury the lead or take too long to matter.

PRM-056write

Evergreen note

When an idea is worth keeping — atomic, linkable, understandable alone in five years.

PRM-031ship

Obviously Awesome positioning

When the product category is crowded and differentiation is fuzzy.

PRM-032ship

PAS conversion section

For a key page section that must convert — hero, pricing block, email.

PRM-033ship

Release notes that build trust

When shipping a release — changelog as credibility, not feature dump.

PRM-051ship

Pricing page audit

When pricing is live or planned but conversion or comprehension is weak.

PRM-052ship

StoryBrand brand script

When marketing talks about the product instead of the customer's story.

PRM-057ship

Launch checklist

When a release is days away — sequence assets, channels, and rollback before you ship.

PRM-058ship

Case study structure

When you need proof that real customers got real results — not marketing fluff.