# Refyne source-reviewed prompt library

Source: https://refyneit.com/prompts

These are third-party prompts with source review, not proof of benchmark performance. Adapt them to your task. Source credit and license links follow each prompt.

# Website content studio

Source: https://refyneit.com/prompts/website-content-studio

A polished section-by-section workspace for planning, drafting, and reviewing a personal site.

## When to use it

It defines hierarchy, states, interaction, content quality, and verification—enough direction for a coding agent to produce a coherent product surface instead of a static mockup.

## Prompt

Build a polished website content studio for [AUDIENCE] inside this app.

First inspect the existing design system, routing, data patterns, and repository guidance. Reuse established components and tokens.

Create a responsive three-part workspace:
- a compact project rail showing site sections and completion status;
- a focused canvas for the selected section with goal, draft, notes, and preview modes;
- a progress area that makes the next useful action obvious.

Use realistic seed content for [SITE OR BRAND], not lorem ipsum. Establish an editorial visual language with calm neutrals, one restrained accent, strong typographic hierarchy, generous space, and crisp dividers. Preserve context while switching sections. Include empty, saving, saved, and validation-error states; keyboard navigation; visible focus; and accessible labels.

Implement meaningful interactions: add/reorder sections, edit a draft, mark a section ready, and switch between draft and preview. Keep controls quiet until relevant. Avoid copying source branding or exact content.

Verify the finished route at desktop and mobile widths. Confirm no horizontal overflow, keyboard reachability, and working state transitions. Finish with the route, files changed, checks run, and one screenshot of each viewport.

## Adaptation and provenance

Adapted from Site Studio's planning-and-drafting canvas. The source preview is structural inspiration; branding and copy should remain original.

Contributor: [@ayangupt](https://github.com/ayangupt)
Repository: https://github.com/github/awesome-copilot
Pinned commit: b95e24caae4f25c5985f8527fd092ccfed039c36
License: [MIT](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/LICENSE)
- [extension](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/site-studio/extension.mjs)
- [preview](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/site-studio/assets/preview.png)
Review: output-reviewed, 2026-08-23. Reviewed the commit-pinned implementation and its committed preview image. The preview is representative source output, not a claim that this adapted prompt reproduces it pixel-for-pixel.
Reviewed output: https://refyneit.com/prompt-outputs/site-studio.png
Output source: https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/site-studio/assets/preview.png
Representative output committed by @ayangupt in github/awesome-copilot, reviewed at b95e24ca.

# Customer feedback themes

Source: https://refyneit.com/prompts/feedback-theme-dashboard

Turn scattered feedback into a clear, calm dashboard of ranked themes and supporting evidence.

## When to use it

The prompt starts from the decision the dashboard supports, then constrains hierarchy, density, semantics, and responsive behavior.

## Prompt

Design and implement a customer-feedback themes dashboard for [PRODUCT OR TEAM]. Inspect the current app conventions before editing.

The page should help someone answer: what matters most, why, and what should we inspect next? Include a compact summary, ranked theme rows or cards, impact and trend signals, source counts, and a detail view with representative feedback. Use realistic, concise seed data.

Favor an analytical editorial style: high information density without clutter, neutral surfaces, one semantic accent, aligned numbers, readable labels, and subtle separators. Do not rely on color alone. Make the highest-impact theme visually obvious while preserving scanability.

Implement search, impact sorting, a time-range filter, theme selection, and empty/loading/error states. On narrow screens, preserve the ranking and core evidence instead of squeezing desktop columns. Use semantic tables or lists where appropriate, accessible controls, visible focus, and reduced-motion-friendly transitions.

Do not copy source branding or text. Use the source output only as a quality and hierarchy reference.

Test the interactions and render at [DESKTOP WIDTH] and [MOBILE WIDTH]. Report changed files, validation commands, and screenshots of the overview plus one selected theme.

## Adaptation and provenance

Adapted from Feedback Themes' grouped customer-signal interface and committed preview.

Contributor: [@jamesmontemagno](https://github.com/jamesmontemagno)
Repository: https://github.com/github/awesome-copilot
Pinned commit: b95e24caae4f25c5985f8527fd092ccfed039c36
License: [MIT](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/LICENSE)
- [extension](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/feedback-themes/extension.mjs)
- [preview](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/feedback-themes/assets/preview.png)
Review: output-reviewed, 2026-08-23. Reviewed the commit-pinned implementation and its committed preview image. The preview is representative source output, not a claim that this adapted prompt reproduces it pixel-for-pixel.
Reviewed output: https://refyneit.com/prompt-outputs/feedback-themes.png
Output source: https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/feedback-themes/assets/preview.png
Representative output committed by @jamesmontemagno in github/awesome-copilot, reviewed at b95e24ca.

# Developer work command center

Source: https://refyneit.com/prompts/developer-work-command-center

A focused dark workspace that brings repository health, priorities, and active work into one view.

## When to use it

It prevents the usual dashboard-card soup by tying every region to a developer decision and specifying a controlled visual system.

## Prompt

Build a developer work command center for [REPOSITORIES OR TEAM]. Read the project guidance and reuse its components, icons, tokens, and data patterns.

Create a responsive dashboard with: a concise greeting and current focus; repository health cards; active pull requests and issues; a short priority queue; recent signals; and a small quick-actions area. Use believable data and clear timestamps. Every module must answer a concrete question or enable an action.

Use a refined dark visual system with layered surfaces, restrained borders, one cool accent, semantic status colors, and strong type contrast. Avoid neon overload, generic gradients, excessive glass effects, and a wall of identical cards. Vary composition according to importance.

Implement repository switching, filtering, expandable detail, loading/empty/error states, and at least one useful quick action. Preserve readable density at laptop widths and collapse to a deliberate single-column mobile flow. Ensure focus indicators, accessible names, contrast, and non-color status cues.

Keep all branding original. Verify real interaction paths, console cleanliness, and layout at desktop and mobile sizes. Return the route, files changed, test results, and screenshots.

## Adaptation and provenance

Adapted from Work Hub's cross-repository developer canvas and its committed dark-mode preview.

Contributor: [@jamesmontemagno](https://github.com/jamesmontemagno)
Repository: https://github.com/github/awesome-copilot
Pinned commit: b95e24caae4f25c5985f8527fd092ccfed039c36
License: [MIT](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/LICENSE)
- [extension](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/work-hub/extension.mjs)
- [preview](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/work-hub/assets/preview.png)
Review: output-reviewed, 2026-08-23. Reviewed the commit-pinned implementation and its committed preview image. The preview is representative source output, not a claim that this adapted prompt reproduces it pixel-for-pixel.
Reviewed output: https://refyneit.com/prompt-outputs/work-hub.png
Output source: https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/work-hub/assets/preview.png
Representative output committed by @jamesmontemagno in github/awesome-copilot, reviewed at b95e24ca.

# Release notes showcase

Source: https://refyneit.com/prompts/release-notes-showcase

Transform a release into a launch-ready story with highlights, changes, contributors, and export controls.

## When to use it

It treats release notes as content design, not a changelog dump, while making accessibility and real product behavior part of the deliverable.

## Prompt

Create a launch-ready release notes showcase for [PRODUCT] version [VERSION]. Inspect the existing app and build it as a real route using current conventions.

Structure the page around a strong release title and summary, two or three visual highlights, grouped changes, contributor recognition, and export/share actions. Write realistic seed copy that distinguishes user value from implementation detail. Make breaking changes or upgrade steps impossible to miss.

Choose a confident, celebratory art direction with original colors, large display type, intentional asymmetry, and a few simple decorative shapes. Keep body content highly readable and avoid visual noise, stock imagery, copied branding, or gratuitous gradients. The result should feel authored, not templated.

Implement category filtering, anchored section navigation, copy/share feedback, and print/export-friendly styling. Include loading, missing-release, and long-content behavior. On mobile, retain the story hierarchy and place controls within thumb reach. Meet keyboard, focus, contrast, and reduced-motion requirements.

Verify the final page with realistic content at desktop and mobile widths. Report files, checks, and screenshots of both views.

## Adaptation and provenance

Adapted from Release Notes Showcase's launch-oriented canvas and committed preview; the new implementation must use original brand expression.

Contributor: [@jamesmontemagno](https://github.com/jamesmontemagno)
Repository: https://github.com/github/awesome-copilot
Pinned commit: b95e24caae4f25c5985f8527fd092ccfed039c36
License: [MIT](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/LICENSE)
- [extension](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/release-notes-showcase/extension.mjs)
- [preview](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/release-notes-showcase/assets/preview.png)
Review: output-reviewed, 2026-08-23. Reviewed the commit-pinned implementation and its committed preview image. The preview is representative source output, not a claim that this adapted prompt reproduces it pixel-for-pixel.
Reviewed output: https://refyneit.com/prompt-outputs/release-notes-showcase.png
Output source: https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/release-notes-showcase/assets/preview.png
Representative output committed by @jamesmontemagno in github/awesome-copilot, reviewed at b95e24ca.

# Swipe-based backlog triage

Source: https://refyneit.com/prompts/swipe-backlog-triage

A fast, card-based flow for making one clear issue decision at a time—with undo and keyboard parity.

## When to use it

The prompt balances expressive interaction with safety, accessibility, and enough issue context to prevent careless triage.

## Prompt

Design and implement a backlog triage flow for [PROJECT] that presents one issue at a time and supports decisions: assign, needs information, defer, close, or ignore.

Inspect existing patterns first. Build a centered decision card with issue title, metadata, concise body, labels, age, and author; a visible queue/progress indicator; and clearly labeled actions. Use realistic issue data. Show enough context for a responsible decision without turning the card into a full issue page.

Create an energetic but disciplined visual style with tactile depth, readable typography, restrained semantic color, and clear spatial cues. Swipe/drag may enhance the experience, but every action must also work by button and keyboard. Add confirmation for destructive choices, undo for the last action, and clear success, empty-queue, loading, and failure states.

Motion should explain the decision and honor reduced-motion preferences. Make mobile interaction comfortable and desktop keyboard use fast. Avoid copied branding and never encode meaning by color alone.

Test all decision paths, undo, keyboard controls, and responsive layout. Finish with changed files, commands/results, and screenshots of the active card and completed state.

## Adaptation and provenance

Adapted from Backlog Swipe Triage's card-based GitHub issue workflow and committed preview.

Contributor: [@jamesmontemagno](https://github.com/jamesmontemagno)
Repository: https://github.com/github/awesome-copilot
Pinned commit: b95e24caae4f25c5985f8527fd092ccfed039c36
License: [MIT](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/LICENSE)
- [extension](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/backlog-swipe-triage/extension.mjs)
- [preview](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/backlog-swipe-triage/assets/preview.png)
Review: output-reviewed, 2026-08-23. Reviewed the commit-pinned implementation and its committed preview image. The preview is representative source output, not a claim that this adapted prompt reproduces it pixel-for-pixel.
Reviewed output: https://refyneit.com/prompt-outputs/backlog-swipe-triage.png
Output source: https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/backlog-swipe-triage/assets/preview.png
Representative output committed by @jamesmontemagno in github/awesome-copilot, reviewed at b95e24ca.

# CI artifact explorer

Source: https://refyneit.com/prompts/ci-artifact-explorer

A trustworthy interface for browsing test reports, recordings, logs, and build artifacts without losing context.

## When to use it

It combines information architecture with an explicit trust boundary, which is essential for an artifact viewer that handles varied, potentially unsafe files.

## Prompt

Build a CI artifact explorer for [PROJECT OR PROVIDER]. Follow repository guidance and existing UI/data conventions.

Create a split workspace: a searchable run/artifact tree on the left and a context-aware viewer on the right. Support test summaries, text logs, screenshots, videos, and safe download metadata. Add breadcrumbs, run status, branch/commit context, file size, timestamps, and a clear path back to the originating job.

Use a quiet technical aesthetic: clean light surfaces, compact but readable rows, monospaced metadata where useful, restrained status colors, and clear selection. Prioritize trust and orientation over decoration.

Implement artifact/run selection, search, filtering, expandable test suites, loading/empty/expired/error states, and safe handling for unsupported or untrusted content. Do not execute embedded artifact code. On mobile, turn the split view into an explicit browse → inspect flow. Ensure semantic tree/list behavior, keyboard navigation, visible focus, and non-color status labels.

Use realistic seed artifacts and original branding. Verify the main flow and responsive layouts, then report files, checks, security assumptions, and screenshots of the browser and selected result.

## Adaptation and provenance

Adapted from PR Artifact Explorer's test-result and workflow-artifact canvas plus its committed preview.

Contributor: [@IEvangelist](https://github.com/IEvangelist)
Repository: https://github.com/github/awesome-copilot
Pinned commit: b95e24caae4f25c5985f8527fd092ccfed039c36
License: [MIT](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/LICENSE)
- [extension](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/pr-artifact-explorer/extension.mjs)
- [preview](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/pr-artifact-explorer/assets/preview.png)
Review: output-reviewed, 2026-08-23. Reviewed the commit-pinned implementation and its committed preview image. The preview is representative source output, not a claim that this adapted prompt reproduces it pixel-for-pixel.
Reviewed output: https://refyneit.com/prompt-outputs/pr-artifact-explorer.png
Output source: https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/pr-artifact-explorer/assets/preview.png
Representative output committed by @IEvangelist in github/awesome-copilot, reviewed at b95e24ca.

# Daily focus board

Source: https://refyneit.com/prompts/daily-focus-board

A calm daily workspace that separates the few priorities that matter from everything else.

## When to use it

The scope and visual constraints keep the experience centered on focus rather than accumulating generic productivity features.

## Prompt

Create a daily focus board for [ROLE OR TEAM] that helps a person choose and complete a small number of meaningful priorities.

Inspect the app's existing design system and persistence patterns. Include today's top outcomes, a lightweight task queue, time or energy context, progress, and a small reflection area. Use realistic content and make the next action unmistakable. Avoid turning it into a full project-management suite.

Design a calm dark interface with warm typography, layered but subtle surfaces, one restrained accent, generous whitespace, and clear completed/active states. Vary panel scale by importance instead of using a uniform card grid. Do not use copied branding, neon gradients, or decorative clutter.

Implement add/edit/complete/reorder, keyboard-friendly controls, optimistic feedback with recovery, an empty first-run state, and a completed-day state. Preserve focus and reading order on mobile. Meet contrast, accessible-name, visible-focus, and reduced-motion expectations.

Persist data using the project's existing approach or a clearly isolated local fallback. Verify interactions, reload behavior, desktop/mobile layouts, and console output. Report route, files, checks, and screenshots.

## Adaptation and provenance

Adapted from Daily Focus Board's prioritized daily-work canvas and committed preview.

Contributor: [@jennyf19](https://github.com/jennyf19)
Repository: https://github.com/github/awesome-copilot
Pinned commit: b95e24caae4f25c5985f8527fd092ccfed039c36
License: [MIT](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/LICENSE)
- [extension](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/daily-focus-board/extension.mjs)
- [preview](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/daily-focus-board/assets/preview.png)
Review: output-reviewed, 2026-08-23. Reviewed the commit-pinned implementation and its committed preview image. The preview is representative source output, not a claim that this adapted prompt reproduces it pixel-for-pixel.
Reviewed output: https://refyneit.com/prompt-outputs/daily-focus-board.png
Output source: https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/daily-focus-board/assets/preview.png
Representative output committed by @jennyf19 in github/awesome-copilot, reviewed at b95e24ca.

# Repository actions dashboard

Source: https://refyneit.com/prompts/repository-actions-dashboard

A workflow operations surface for browsing runs, spotting failures, and safely dispatching an action.

## When to use it

It gives a coding agent the operational hierarchy, safe action flow, and edge states needed for a credible workflow UI.

## Prompt

Build a repository actions dashboard for [REPOSITORY OR AUTOMATION SYSTEM]. Start by studying repository guidance and existing components.

Create an operations-focused layout with a workflow rail, recent run list, status/duration/branch/actor metadata, a selected-run detail panel, and a deliberate “run workflow” action. Use believable seed data and clear relative plus absolute timestamps.

Use a refined dark theme with compact rows, strong alignment, restrained borders, and semantic status treatment that includes icons and text. Give failures and actionable states priority without making the whole page alarming. Avoid decorative dashboards and repeated cards where a list is more scannable.

Implement workflow selection, status/branch filtering, run detail, log-step expansion, refresh feedback, and a dispatch dialog with validation and confirmation. Include loading, no-runs, permission-denied, rate-limit, and failure states. Make dense desktop use keyboard-efficient; use a browse → detail flow on mobile.

Do not copy source branding. Verify filters, dialog focus, dispatch safety, responsive layouts, and accessibility. Report changed files, checks, and screenshots of overview and run detail.

## Adaptation and provenance

Adapted from Repo Actions Hub's workflow browsing and dispatch canvas with its committed preview.

Contributor: [@jamesmontemagno](https://github.com/jamesmontemagno)
Repository: https://github.com/github/awesome-copilot
Pinned commit: b95e24caae4f25c5985f8527fd092ccfed039c36
License: [MIT](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/LICENSE)
- [extension](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/repo-actions-hub/extension.mjs)
- [preview](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/repo-actions-hub/assets/preview.png)
Review: output-reviewed, 2026-08-23. Reviewed the commit-pinned implementation and its committed preview image. The preview is representative source output, not a claim that this adapted prompt reproduces it pixel-for-pixel.
Reviewed output: https://refyneit.com/prompt-outputs/repo-actions-hub.png
Output source: https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/repo-actions-hub/assets/preview.png
Representative output committed by @jamesmontemagno in github/awesome-copilot, reviewed at b95e24ca.

# Modernization workflow studio

Source: https://refyneit.com/prompts/modernization-workflow-studio

Guide a complex migration from readiness through assessment, prioritized work, progress, and validation.

## When to use it

It translates a complex technical process into visible stages, gates, evidence, and next actions—the core of a useful workflow product.

## Prompt

Design and implement a modernization workflow studio for [SYSTEM OR MIGRATION]. Reuse the project's architecture and design system.

Model a clear multi-stage journey: environment readiness, assessment, prioritized plan, execution progress, and validation. Provide a persistent stage navigator, current-stage guidance, evidence/status summaries, blockers, and the next safe action. Use realistic findings and tasks rather than placeholder text.

Choose a professional light visual system with strong hierarchy, calm neutral surfaces, a single primary accent, semantic statuses with text/icons, and progressive disclosure for technical detail. The interface should reduce uncertainty during a long-running effort, not resemble a generic admin panel.

Implement stage navigation, expandable findings, task state changes, progress calculation, validation results, loading/empty/error/blocked states, and persistence using existing project patterns. Prevent skipping required gates without explanation. On mobile, retain progress and next action while collapsing secondary evidence deliberately.

Keep branding and content original. Verify gating logic, keyboard flow, responsive rendering, and state recovery. Finish with files changed, commands/results, assumptions, and screenshots of overview and one detail stage.

## Adaptation and provenance

Adapted from Java Modernization Studio's readiness-to-validation workflow and committed preview; generalized beyond Java.

Contributor: [@ayangupt](https://github.com/ayangupt)
Repository: https://github.com/github/awesome-copilot
Pinned commit: b95e24caae4f25c5985f8527fd092ccfed039c36
License: [MIT](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/LICENSE)
- [extension](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/java-modernization-studio/extension.mjs)
- [preview](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/java-modernization-studio/assets/preview.png)
Review: output-reviewed, 2026-08-23. Reviewed the commit-pinned implementation and its committed preview image. The preview is representative source output, not a claim that this adapted prompt reproduces it pixel-for-pixel.
Reviewed output: https://refyneit.com/prompt-outputs/java-modernization-studio.png
Output source: https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/java-modernization-studio/assets/preview.png
Representative output committed by @ayangupt in github/awesome-copilot, reviewed at b95e24ca.

# Submission readiness workshop

Source: https://refyneit.com/prompts/submission-readiness-workshop

A guided workspace that turns eligibility checks and remediation into a confident final submission.

## When to use it

It makes readiness explainable and actionable by connecting requirements, evidence, remediation, gating, and progress in one flow.

## Prompt

Build a submission readiness workshop for [PROGRAM, MARKETPLACE, OR REVIEW PROCESS]. Inspect the codebase and use its current patterns.

Create a guided workspace with eligibility summary, weighted readiness score, grouped requirements, evidence links, remediation tasks, and a final submission step. Make blocking versus recommended work unmistakable. Use realistic criteria and concise helper copy.

Use a polished workshop aesthetic: dark layered surfaces, one warm accent, strong section numbering, clear progress, and small moments of encouragement. Keep information readable and serious; avoid gamification that obscures risk, copied branding, or excessive effects.

Implement requirement expansion, evidence attachment or link entry, completion tracking, score updates, save/resume, and a final review confirmation. Include first-run, partial, ready, validation-error, and submission-failure states. Destructive resets need confirmation. Ensure keyboard access, visible focus, contrast, non-color status cues, and a deliberate mobile step flow.

Verify persistence, score logic, blocked submission, successful completion, and responsive layouts. Return the route, files changed, checks, and screenshots of the in-progress and ready states.

## Adaptation and provenance

Adapted from Tiny Tool Town Submitter's inspect-remediate-submit workshop and committed preview; generalized to other review processes.

Contributor: [@jamesmontemagno](https://github.com/jamesmontemagno)
Repository: https://github.com/github/awesome-copilot
Pinned commit: b95e24caae4f25c5985f8527fd092ccfed039c36
License: [MIT](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/LICENSE)
- [extension](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/tiny-tool-town-submitter/extension.mjs)
- [preview](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/tiny-tool-town-submitter/assets/preview.png)
Review: output-reviewed, 2026-08-23. Reviewed the commit-pinned implementation and its committed preview image. The preview is representative source output, not a claim that this adapted prompt reproduces it pixel-for-pixel.
Reviewed output: https://refyneit.com/prompt-outputs/tiny-tool-town-submitter.png
Output source: https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/extensions/tiny-tool-town-submitter/assets/preview.png
Representative output committed by @jamesmontemagno in github/awesome-copilot, reviewed at b95e24ca.

# Reproduce → diagnose → fix

Source: https://refyneit.com/prompts/reproduce-diagnose-fix

Debug from a confirmed reproduction and competing hypotheses instead of guessing at a patch.

## When to use it

The prompt makes reproduction and falsifiable hypotheses gates before editing, so a plausible patch is not mistaken for a demonstrated fix.

## Prompt

Debug [BUG OR FAILURE] in this repository.

Work in this order:
1. Read the relevant repository guidance, error output, recent changes, and nearby tests.
2. Reproduce the failure before editing. Record exact steps, expected behavior, actual behavior, environment details, and the smallest useful error/log excerpt. If reproduction is impossible, say why and gather the strongest available evidence instead.
3. Trace the affected execution path and form at least two specific root-cause hypotheses. Rank them and define a falsifying check for each.
4. Test the hypotheses. Do not suppress errors, weaken tests, or add a broad fallback as a substitute for understanding the cause.
5. Implement the smallest fix that addresses the confirmed cause and follows existing project patterns.
6. Add or update a regression test that fails before the fix and passes after it.
7. Run the targeted test, the original reproduction, and the relevant broader checks.

Finish with:
- Root cause
- Fix and files changed
- Proof, including commands and results
- Remaining risks or adjacent issues you deliberately did not change

## Adaptation and provenance

Condensed and made tool-neutral from GitHub Awesome Copilot's Debug Mode Instructions.

Contributor: [@aaronpowell](https://github.com/aaronpowell)
Repository: https://github.com/github/awesome-copilot
Pinned commit: b95e24caae4f25c5985f8527fd092ccfed039c36
License: [MIT](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/LICENSE)
- [debug](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/agents/debug.agent.md)
Review: source-reviewed, 2026-08-23. Compared with the commit-pinned source and adapted into a tool-neutral, copy/paste request. This is a non-visual task, so no output artifact is required.

# TDD: red → green → refactor

Source: https://refyneit.com/prompts/tdd-red-green-refactor

Deliver one behavior at a time with visible failing and passing test evidence.

## When to use it

Each requested behavior gets an observable red and green signal, while the refactor boundary prevents speculative cleanup from hiding inside the feature diff.

## Prompt

Implement [BEHAVIOR OR ISSUE] using strict red → green → refactor cycles.

First, read the issue/specification, repository guidance, existing implementation, and nearby tests. Convert the requested behavior into a short checklist of externally observable outcomes.

For each outcome, one at a time:
1. RED — write the smallest meaningful test for that behavior. Run it and confirm it fails for the expected behavioral reason, not because of setup, import, or syntax errors. Show the test name and concise failure.
2. GREEN — make the minimum production change needed to pass that test. Run the targeted test and show the passing result.
3. REFACTOR — improve names or structure only while the tests remain green. Do not add unrelated abstractions or features.

After all cycles:
- Run the relevant full suite and project checks.
- Cross-check every acceptance criterion against a test or an explicit verification step.
- Report cycles completed, behaviors covered, commands/results, and anything intentionally left untested with the reason.

Never modify a test merely to make an incorrect implementation pass. If the specification conflicts with existing behavior, stop and describe the conflict before choosing a side.

## Adaptation and provenance

Combined and condensed from the red, green, and refactor agents in GitHub Awesome Copilot.

Contributor: [@aaronpowell](https://github.com/aaronpowell)
Repository: https://github.com/github/awesome-copilot
Pinned commit: b95e24caae4f25c5985f8527fd092ccfed039c36
License: [MIT](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/LICENSE)
- [tdd-red](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/agents/tdd-red.agent.md)
- [tdd-green](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/agents/tdd-green.agent.md)
- [tdd-refactor](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/agents/tdd-refactor.agent.md)
Review: source-reviewed, 2026-08-23. Compared with the commit-pinned source and adapted into a tool-neutral, copy/paste request. This is a non-visual task, so no output artifact is required.

# Security review with evidence

Source: https://refyneit.com/prompts/security-review-evidence

Review an actual change for exploitable trust-boundary failures and rank findings by risk.

## When to use it

A trust-boundary inventory plus required abuse paths filters out generic scanner prose and makes each finding reproducible and triageable.

## Prompt

Perform a security review of [CHANGE, PR, OR COMPONENT]. Do not edit code.

Establish scope first: identify entry points, data stores, external services, identities, privileges, secrets, and trust boundaries. Read the relevant diff plus the surrounding code that controls authentication, authorization, validation, serialization, persistence, logging, and error handling.

Test the change against these risk areas where applicable:
- authentication and session handling
- resource-level and function-level authorization
- injection, unsafe deserialization, SSRF, path traversal, and command execution
- secrets or sensitive data exposure in source, logs, errors, telemetry, caches, or clients
- insecure defaults, missing rate limits, replay/race conditions, and dependency risk
- prompt injection, untrusted tool output, and excessive agency for AI features

For every finding provide:
1. Severity: critical / high / medium / low
2. Exact file and line or symbol
3. Attacker prerequisites and a concrete abuse path
4. Impact and affected data or capability
5. Evidence from the code; clearly label anything inferred
6. The smallest safe remediation
7. A verification test that would prove the fix

Do not report style issues as vulnerabilities. Do not claim a control is missing until you trace callers and middleware. End with reviewed surfaces, clean controls you verified, unresolved unknowns, and an overall risk assessment.

## Adaptation and provenance

Condensed and made repository-review focused from GitHub Awesome Copilot's SE: Security agent.

Contributor: [@niksacdev](https://github.com/niksacdev)
Repository: https://github.com/github/awesome-copilot
Pinned commit: b95e24caae4f25c5985f8527fd092ccfed039c36
License: [MIT](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/LICENSE)
- [se-security-reviewer](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/agents/se-security-reviewer.agent.md)
Review: source-reviewed, 2026-08-23. Compared with the commit-pinned source and adapted into a tool-neutral, copy/paste request. This is a non-visual task, so no output artifact is required.

# Runtime accessibility audit

Source: https://refyneit.com/prompts/accessibility-runtime-audit

Exercise real keyboard, focus, form, dialog, zoom, and screen-reader paths with evidence.

## When to use it

It centers the audit on complete user tasks and manual runtime behavior, which catches focus and announcement failures that static checks routinely miss.

## Prompt

Audit [URL OR USER FLOW] for runtime accessibility. Do not rely on source inspection or an automated score alone.

Before testing, identify the critical tasks a user must complete and the browser, viewport, and assistive-technology setup available.

Exercise each critical task with:
1. Keyboard only — logical tab order, visible focus, no traps, usable menus/dialogs, and alternatives to pointer-only gestures.
2. Focus management — initial focus, focus containment where appropriate, escape behavior, and restoration to the trigger.
3. Forms — programmatic labels, instructions, autocomplete, retained input, understandable inline errors, and error-summary behavior.
4. Dynamic updates — route changes, loading, validation, and success/error states announced at the right time without duplicate noise.
5. Reflow and perception — 200% and 400% zoom, narrow viewport, text spacing, reduced motion, high contrast, and color-independent meaning.
6. Automated checks — run the project's existing accessibility tools, then manually validate every reported blocker.

For each finding include severity, affected task, WCAG success criterion when known, exact reproduction steps, observed vs expected behavior, evidence, and a testable fix recommendation. Separate confirmed failures from risks you could not test.

Finish with tested paths, environment, automated-tool results, manual findings, verified clean behaviors, and coverage gaps.

## Adaptation and provenance

Condensed and made tool-neutral from GitHub Awesome Copilot's Accessibility Runtime Tester.

Contributor: [@proflead](https://github.com/proflead)
Repository: https://github.com/github/awesome-copilot
Pinned commit: b95e24caae4f25c5985f8527fd092ccfed039c36
License: [MIT](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/LICENSE)
- [accessibility-runtime-tester](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/agents/accessibility-runtime-tester.agent.md)
Review: source-reviewed, 2026-08-23. Compared with the commit-pinned source and adapted into a tool-neutral, copy/paste request. This is a non-visual task, so no output artifact is required.

# Technical-debt remediation plan

Source: https://refyneit.com/prompts/technical-debt-remediation

Turn a vague cleanup request into evidence-ranked debt with bounded, verifiable work packages.

## When to use it

The evidence and cost-of-delay requirements keep the plan tied to product risk rather than aesthetic preference, while work packages make the result executable.

## Prompt

Assess technical debt in [SCOPE] and produce a remediation plan. Do not edit code.

Build an evidence inventory from code, tests, build output, dependency reports, incidents, performance data, and existing TODOs. Do not label unfamiliar code as debt merely because you prefer another style.

For each candidate debt item record:
- exact location and current behavior
- evidence that it creates defects, slows delivery, increases security/operational risk, or blocks a stated goal
- impact and likelihood
- dependencies and affected owners/surfaces
- estimated remediation size: S / M / L / XL
- cost of delay and what happens if it remains unchanged

Rank items with a stated method using impact, urgency, confidence, and effort. Group the selected items into independently shippable work packages. Each package must include scope, non-goals, prerequisites, implementation outline, migration/rollback concerns, acceptance criteria, verification commands, and observability needed after release.

Separate:
1. Fix now
2. Schedule with a trigger or deadline
3. Monitor with a named signal
4. Accept intentionally, with rationale

End with the top three recommended actions, the evidence behind their order, quick wins, high-risk unknowns, and the next review date.

## Adaptation and provenance

Condensed and made tool-neutral from GitHub Awesome Copilot's Tech Debt Remediation Plan agent.

Contributor: [@aaronpowell](https://github.com/aaronpowell)
Repository: https://github.com/github/awesome-copilot
Pinned commit: b95e24caae4f25c5985f8527fd092ccfed039c36
License: [MIT](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/LICENSE)
- [tech-debt-remediation-plan](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/agents/tech-debt-remediation-plan.agent.md)
Review: source-reviewed, 2026-08-23. Compared with the commit-pinned source and adapted into a tool-neutral, copy/paste request. This is a non-visual task, so no output artifact is required.

# Repository-grounded implementation plan

Source: https://refyneit.com/prompts/implementation-plan-repo-grounded

Produce an executable plan tied to real files, contracts, dependencies, and verification.

## When to use it

The plan must prove it understands the current system and connect every task to files, contracts, and verification, making it useful to the next engineer or agent.

## Prompt

Create an implementation plan for [FEATURE OR REFACTOR]. Do not edit code.

Start by reading repository guidance and tracing the current behavior end to end. Identify the relevant entry points, callers, data contracts, persistence, external integrations, tests, build/deploy configuration, and existing patterns that the change should reuse.

State:
- the problem and desired observable outcome
- confirmed current behavior with file/symbol references
- assumptions and unresolved questions
- in-scope and explicitly out-of-scope work
- compatibility, migration, security, privacy, and data-loss constraints

Then break the implementation into the smallest independently verifiable phases. For every task include:
1. Goal and rationale
2. Exact files and symbols to change or add
3. Behavior and contract changes
4. Dependencies on earlier tasks
5. Tests to add or update
6. Verification command or manual check
7. Rollback or failure considerations when relevant

Sequence tasks according to real dependencies. Call out tasks that can run in parallel. Do not invent file names, APIs, or architecture not supported by the repository; label proposals as proposals.

Finish with a requirement-to-task traceability table, release/rollout steps, observability, and a definition-of-done checklist.

## Adaptation and provenance

Condensed and relaxed from GitHub Awesome Copilot's Implementation Plan Generation Mode so it fits existing repository conventions.

Contributor: [@aaronpowell](https://github.com/aaronpowell)
Repository: https://github.com/github/awesome-copilot
Pinned commit: b95e24caae4f25c5985f8527fd092ccfed039c36
License: [MIT](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/LICENSE)
- [implementation-plan](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/agents/implementation-plan.agent.md)
Review: source-reviewed, 2026-08-23. Compared with the commit-pinned source and adapted into a tool-neutral, copy/paste request. This is a non-visual task, so no output artifact is required.

# Product requirements document

Source: https://refyneit.com/prompts/product-requirements-document

Turn a product idea into testable outcomes, non-goals, stories, metrics, and rollout constraints.

## When to use it

Traceability from goals through metrics prevents the PRD from becoming a feature wish list and makes later scope and launch decisions inspectable.

## Prompt

Write a product requirements document for [PRODUCT OR FEATURE].

Use the supplied research, user feedback, analytics, and technical context as evidence. If evidence is missing, ask only questions that materially change scope; otherwise state assumptions and continue. Never invent customer quotes, market facts, or metrics.

Include:
1. Executive summary and the user problem
2. Target users and jobs to be done
3. Evidence and current baseline
4. Goals as measurable outcomes, plus non-goals
5. User journeys and prioritized user stories
6. Functional requirements with unique IDs
7. Non-functional requirements: accessibility, performance, privacy, security, reliability, and supportability
8. Dependencies, constraints, edge cases, and failure states
9. Analytics events, success metrics, guardrail metrics, and decision thresholds
10. Rollout stages, experiment or validation plan, rollback triggers, and support readiness
11. Risks, open questions, and named decision owners where known

Every requirement must have testable acceptance criteria. Use Given/When/Then only where it improves precision. Include a traceability table connecting goals → requirements → acceptance criteria → metrics.

Finish with the smallest viable release, what evidence would justify expansion, and what evidence would cause the team to stop or revise the plan.

## Adaptation and provenance

Condensed and made evidence-first from GitHub Awesome Copilot's Create PRD Chat Mode.

Contributor: [@aaronpowell](https://github.com/aaronpowell)
Repository: https://github.com/github/awesome-copilot
Pinned commit: b95e24caae4f25c5985f8527fd092ccfed039c36
License: [MIT](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/LICENSE)
- [prd](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/agents/prd.agent.md)
Review: source-reviewed, 2026-08-23. Compared with the commit-pinned source and adapted into a tool-neutral, copy/paste request. This is a non-visual task, so no output artifact is required.

# Architecture decision record

Source: https://refyneit.com/prompts/architecture-decision-record

Document a consequential technical choice with steelmanned options and reversal signals.

## When to use it

Ranked drivers and reconsideration triggers turn an ADR from a retrospective justification into a reversible, evidence-aware decision tool.

## Prompt

Create an Architecture Decision Record for [DECISION]. Do not implement the decision.

First inspect the repository and supplied context so the record reflects the actual system. Separate confirmed facts from assumptions.

Write the ADR with:
- Title, date, status, owners, and decision scope
- Context: the forces creating the decision, current constraints, and why a decision is needed now
- Decision drivers ranked by importance
- At least two viable options plus “do nothing” when meaningful
- A steelman of each option, including operational, security, privacy, performance, cost, migration, and team-skill effects
- Decision and concise rationale tied directly to the ranked drivers
- Positive and negative consequences, including new failure modes and long-term maintenance
- Implementation and migration outline, compatibility needs, and rollback/reversal path
- Validation plan with measurable signals and review date
- Reconsideration triggers: the evidence or changed constraints that should reopen the decision
- References to relevant files, issues, benchmarks, incidents, or documentation

Do not fabricate benchmark results or consensus. If the decision cannot responsibly be made from current evidence, set status to Proposed and list the minimum experiments or stakeholder decisions required.

## Adaptation and provenance

Condensed and made repository-grounded from GitHub Awesome Copilot's ADR Generator.

Contributor: [@askpt](https://github.com/askpt)
Repository: https://github.com/github/awesome-copilot
Pinned commit: b95e24caae4f25c5985f8527fd092ccfed039c36
License: [MIT](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/LICENSE)
- [adr-generator](https://github.com/github/awesome-copilot/blob/b95e24caae4f25c5985f8527fd092ccfed039c36/agents/adr-generator.agent.md)
Review: source-reviewed, 2026-08-23. Compared with the commit-pinned source and adapted into a tool-neutral, copy/paste request. This is a non-visual task, so no output artifact is required.