AI Prompt Library

Turns a rough idea of what you want an AI to do into a complete, production ready system prompt.

A senior engineer review pass that cites specific lines, explains the risk, and proposes the fix.

Audits an interface against Nielsen's 10 usability heuristics and ranks the fixes by severity.

Turns a table of numbers into a narrative an audience will remember, without overclaiming causation.

Upscales a reference photo to 4K while holding identity, pose, lighting, and background exactly as they were.

Repairs an old or damaged photo, scratches, fading, and blur included, without inventing faces that were never there.

Generates a multi angle character turnaround from one reference image, holding face, outfit, and art style consistent.

Lays out a clean product sheet with hero render, feature callouts, specs, and colorways from your own product details.

Produces a batch of stock ready image concepts with the generation prompt, title, and keyword set for each.

Expands a one line scene idea into a full pixel art prompt with era, palette, dithering, and readability constraints.

Renders a badly taken photo of a phone screen showing a messaging thread, imperfections and glare included.

A bold screen printed poster look built from repeated motifs, halftone grit, and neon on black.

Builds an editorial travel poster collage of a traveler in a named country, with layered paper texture and a headline.

A worked example of a JSON structured image prompt, specifying subject, wardrobe, scene, and camera separately.

Walks an unfamiliar TypeScript codebase and produces the structured map a new developer actually needs.

A full review pass across correctness, performance, security, tests, and docs, in any language.

Deep analysis of a Git repository into a knowledge base covering architecture, data flow, tests, and deploys.

A heavyweight framework for building production prompts that cite their sources and state their own failure modes.

Audits a TypeScript codebase for weak typing: escaped anys, loose generics, and inference you are fighting.

Scores a change for blast radius, test coverage, dependency risk, and backward compatibility before you merge it.

A structured template for stating a plan's direction of effect, then hunting its loopholes until none survive.

Reviews a requirement from seven separate roles in turn, so the gaps one perspective misses another one catches.

Specifies match 3 grid logic properly: detection, cascades, scoring, and special piece rules that do not conflict.

Builds 2D vector space combat with momentum, collision, and weapons that behave consistently at any frame rate.

A reproduce, isolate, hypothesize, prove loop that finds the actual cause instead of the first plausible one.

Writes tests that could actually fail, covering the boundaries and error paths rather than restating the code.

Restructures code in verifiable steps, with an explicit account of what behavior must not change.

Explains what unfamiliar code does, why it is shaped that way, and which oddities are load bearing.

A thirteen domain forensic review protocol for TypeScript apps and packages, with a severity ranked report at the end.

Refactors an existing component onto shadcn/ui structure and styling without touching its logic or props.

Forces the agent to explain the whole system flow before it is allowed to change a line, then fix minimally.

The explain before you fix debugging discipline, specialized for Flutter map and GIS layer rendering bugs.

A short instruction block to append to any coding request that pushes back on overengineering.

Attacks a plan for loopholes and unstated assumptions, then loops until a pass turns up nothing new.

Turns your own recent history with a coding agent into five specific workflow changes, ranked by payoff.

A fill in the blanks brief for implementing a feature, with styling, error handling, tests, and docs built in.

Generates a PRD or a technical design doc with the sections stakeholders actually ask about.

Design direction for modern marketing pages: motion, depth, current palettes, and SEO structure together.

Makes the agent write and get agreement on a spec before it generates any files.

A standing rule that makes an agent state its intent and blast radius before each change, not after.

Makes an agent review its own work adversarially and report what it is unsure about before you look at it.

Stops a drifting session, separates the original task from everything that accumulated, and gets you a landable slice.

A React 19 and Next.js App Router specialist that respects the server and client component boundary.

Writes Storybook stories that cover the variants, states, and edge cases you would have forgotten.

Designs a reusable website template system rather than a single page, with tokens, sections, and rules.

A combined architecture and UI review that reports at the level of decisions, not line by line nitpicks.

Strict Astro architecture rules that keep islands, content collections, and rendering modes from drifting.

A component library architect that designs the API and composition model before anyone writes JSX.

An SEO strategist covering keyword research, on page structure, and the technical work that gates the rest.

A technical and on page SEO audit that returns a prioritized remediation list rather than a score.

A frontend implementer that treats responsiveness, accessibility, and performance as requirements, not extras.

A WCAG 2.1 and 2.2 audit covering screen readers, keyboard paths, contrast, and focus management.

Turns a failing Lighthouse or field data report into a ranked fix list with the mechanism behind each number.

Diagnoses why a layout breaks at certain widths and fixes the cause rather than adding another breakpoint.

Reviews an API surface for contract, status codes, validation, authorization, and the failure cases nobody wrote.

Reviews a form for the validation timing, error copy, and failure handling that decide whether people finish it.

Turns a vague request into an implementation plan with scope, risks, and the questions nobody asked yet.

A skill that pulls a social post apart, verifies its claims, and turns it into usable source material.

Runs a ticket end to end: fetch requirements, design, implement, test, and open the pull request.

A readiness audit that asks whether a feature is genuinely done, including the parts nobody demoed.

Traces data lineage across scripts and stored procedures to show where a field actually comes from.

A disciplined debugging loop that isolates the cause before changing anything, then proves the fix.

Hardens AI chat and completion endpoints against prompt injection, PII leakage, and abuse.

Compresses a whole technique into a system prompt under a hard character budget, through scored iterations.

Maintains a PROGRESS.md as durable working memory so an agent survives losing its context window.

Wires an agent into Trello to list boards, read lists, and create cards from what it is working on.

A skill that gets an agent productive in an unfamiliar repository by reading it in the right order.

Turns a commit range into a changelog written for users, sorted by what they will notice.

Researches a question and returns a brief that separates what is established from what is contested or unknown.

Writes a PR description from the actual diff, aimed at whoever has to review it and whoever finds it in a year.

Diagnoses why a prompt is underperforming and rewrites it, showing what changed and why each change matters.

Designs a small set of examples that teach a model the pattern, including the edge cases it would otherwise miss.

Builds a scoring rubric for LLM output that two different reviewers would apply the same way.

Works out why a model keeps ignoring one specific instruction and gives you fixes ranked by how much they cost.