# Agent Kit

> 60 skills and 238 prompts for AI coding agents (Claude Code, Codex, Cursor and any agent that reads Agent Skills). Every page is public and has a markdown twin at the same URL plus .md.

Install: `npx skills add AtlasScope/agent-kit`. Public read-only MCP server: https://atlasscope.dev/mcp. Everything in one file: https://atlasscope.dev/llms-full.txt.

## Kits

- [design-kit](https://atlasscope.dev/kits/design-kit.md): A design library for AI coding agents: a /design router that picks the right skill or recipe, plus vetted skills for typography, color, motion, UX writing, usability, landing copy, native iOS and more.
- [engineering-kit](https://atlasscope.dev/kits/engineering-kit.md): Engineering skills for AI coding agents, vetted and pinned: bug diagnosis, test-first work, architecture scans, plan interviews, retros, primary-source research, setup wizards and read-only PostgreSQL investigations. Every command runs only when you call it.
- [Experience Architecture](https://atlasscope.dev/kits/experience-architecture.md): Find out why a web app's navigation, pages, links or Back button feel wrong, decide whether its structure should change, and prove it in a real browser. Three skills (Diagnose, Rethink, Verify) over a zero-dependency toolkit for route maps, browser journeys, prototypes and URL migrations. Strongest on React and Next.js.
- [working-style](https://atlasscope.dev/kits/working-style.md): Four commands for working with an agent: /ceo explains in plain English, /recommend makes the call, /harsh-review gives direct critique with bounded fixes, /viewport-proof judges a screen across viewports with a proof sheet. Each runs only when you call it.

## Fix a problem

- [Numbers on screen don't match the source](https://atlasscope.dev/fix/numbers-on-screen-dont-match-the-source.md)
- [A workflow got slow, or database cost jumped](https://atlasscope.dev/fix/a-workflow-got-slow-or-database-cost-jumped.md)
- [Tests are green but production keeps breaking](https://atlasscope.dev/fix/tests-are-green-but-production-keeps-breaking.md)
- [Agents break nearby features on small edits](https://atlasscope.dev/fix/agents-break-nearby-features-on-small-edits.md)
- [Things happen twice, or not at all](https://atlasscope.dev/fix/things-happen-twice-or-not-at-all.md)
- [Merged but not live, or a deploy broke something](https://atlasscope.dev/fix/merged-but-not-live-or-a-deploy-broke-something.md)
- [Worried about who can see or change what](https://atlasscope.dev/fix/worried-about-who-can-see-or-change-what.md)
- [The app hangs when an outside service is down](https://atlasscope.dev/fix/the-app-hangs-when-an-outside-service-is-down.md)
- [Your computer is full or browser tests are flaky](https://atlasscope.dev/fix/your-computer-is-full-or-browser-tests-are-flaky.md)
- [Users get lost, or Back and shared links misbehave](https://atlasscope.dev/fix/users-get-lost-or-back-and-shared-links-misbehave.md)
- [The numbers don't match their source](https://atlasscope.dev/fix/the-numbers-dont-match-their-source.md)
- [Records are merged, duplicated or lose history](https://atlasscope.dev/fix/records-are-merged-duplicated-or-lose-history.md)
- [Imports rewrite everything or run twice](https://atlasscope.dev/fix/imports-rewrite-everything-or-run-twice.md)
- [Nobody knows what writes to a table](https://atlasscope.dev/fix/nobody-knows-what-writes-to-a-table.md)
- [Too many views, functions and triggers](https://atlasscope.dev/fix/too-many-views-functions-and-triggers.md)
- [Postgres is slow or costs too much](https://atlasscope.dev/fix/postgres-is-slow-or-costs-too-much.md)
- [A migration is about to ship, or a restore is untested](https://atlasscope.dev/fix/a-migration-is-about-to-ship-or-a-restore-is-untested.md)
- [Agents keep asking me to decide](https://atlasscope.dev/fix/agents-keep-asking-me-to-decide.md)

## Prompt areas

- [Architecture and hidden complexity](https://atlasscope.dev/prompts/architecture-and-hidden-complexity.md): 8 prompts
- [Performance and scale](https://atlasscope.dev/prompts/performance-and-scale.md): 9 prompts
- [Correctness and reliability](https://atlasscope.dev/prompts/correctness-and-reliability.md): 7 prompts
- [Testing and security](https://atlasscope.dev/prompts/testing-and-security.md): 7 prompts
- [Infrastructure and environment](https://atlasscope.dev/prompts/infrastructure-and-environment.md): 7 prompts
- [Data integrity and analytics](https://atlasscope.dev/prompts/data-integrity-and-analytics.md): 6 prompts
- [Agent harness and modern stack](https://atlasscope.dev/prompts/agent-harness-and-modern-stack.md): 6 prompts
- [Start here](https://atlasscope.dev/prompts/start-here.md): 1 prompts
- [Accuracy and evidence](https://atlasscope.dev/prompts/accuracy-and-evidence.md): 4 prompts
- [Writes and ingestion](https://atlasscope.dev/prompts/writes-and-ingestion.md): 5 prompts
- [Schema and dependencies](https://atlasscope.dev/prompts/schema-and-dependencies.md): 3 prompts
- [Performance and maintenance](https://atlasscope.dev/prompts/performance-and-maintenance.md): 2 prompts
- [Rebuild and safe change](https://atlasscope.dev/prompts/rebuild-and-safe-change.md): 2 prompts
- [Agent decisions](https://atlasscope.dev/prompts/agent-decisions.md): 1 prompts
- [Workflows](https://atlasscope.dev/prompts/workflows.md): 5 prompts
- [Thinking prompts](https://atlasscope.dev/prompts/thinking-prompts.md): 11 prompts
- [Design recipes](https://atlasscope.dev/prompts/design-recipes.md): 22 prompts

## Key prompts

- [Status hidden in a pile of flags](https://atlasscope.dev/prompts/status-hidden-in-a-pile-of-flags.md): Finds features whose real status is spread across a pile of yes/no flags, so combinations that should be impossible can happen.
- [Meaning drift across the stack](https://atlasscope.dev/prompts/meaning-drift-across-the-stack.md): Shows where one word (a task, a fund, a run) means slightly different things in different parts of the app.
- [Layers that don't earn their keep](https://atlasscope.dev/prompts/layers-that-dont-earn-their-keep.md): Finds extra layers of code you could delete so the code, and the agents reading it, get simpler.
- [Two systems that disagree](https://atlasscope.dev/prompts/two-systems-that-disagree.md): Finds data kept in two places (a saved copy and a table, the browser and the server) that can quietly disagree.
- [Code that relies on lucky timing](https://atlasscope.dev/prompts/code-that-relies-on-lucky-timing.md): Finds code that only works because things usually happen in the right order.
- [Timers and connections nobody closes](https://atlasscope.dev/prompts/timers-and-connections-nobody-closes.md): Finds timers, listeners and connections that nobody clearly owns or shuts off.
- [Screens that fight how people work](https://atlasscope.dev/prompts/screens-that-fight-how-people-work.md): Finds when confusing screens are a symptom of how the product is organised rather than a list of separate bugs.
- [Slowest journeys, end to end](https://atlasscope.dev/prompts/slowest-journeys-end-to-end.md): Shows where the seconds go on the paths people actually use.
- [Steps that wait for no reason](https://atlasscope.dev/prompts/steps-that-wait-for-no-reason.md): Finds steps that wait for each other to finish when they could run at the same time.
- [Queries that grow badly](https://atlasscope.dev/prompts/queries-that-grow-badly.md): Finds the database queries that will be the first to slow down as tables grow.
- [Work repeated for nothing](https://atlasscope.dev/prompts/work-repeated-for-nothing.md): Finds work redone on every request or screen redraw when nothing it depends on changed.
- [Loops that blow up with big data](https://atlasscope.dev/prompts/loops-that-blow-up-with-big-data.md): Finds loops inside loops that are fine at 50 rows and painful at 50,000.
- [Memory that keeps growing](https://atlasscope.dev/prompts/memory-that-keeps-growing.md): Finds why a page or process gets slower or crashes the longer it runs.
- [One slow step that blocks the rest](https://atlasscope.dev/prompts/one-slow-step-that-blocks-the-rest.md): Finds one slow step that makes every other request wait in line behind it.
- [Parallel work with no limit](https://atlasscope.dev/prompts/parallel-work-with-no-limit.md): Finds work run in parallel with no limit, which can swamp the database connections or an outside service.
- [Safe to run twice?](https://atlasscope.dev/prompts/safe-to-run-twice.md): Finds actions that do damage when they run twice: double sends, double charges, duplicate rows.
- [Fallbacks that look like real data](https://atlasscope.dev/prompts/fallbacks-that-look-like-real-data.md): Finds screens that show a confident zero or an empty list when the real answer is 'we could not load this'.
- [Rules the data never enforces](https://atlasscope.dev/prompts/rules-the-data-never-enforces.md): Finds the rules everyone assumes (one owner per task, no negative balance) that nothing actually checks.
- [Updates that overwrite each other](https://atlasscope.dev/prompts/updates-that-overwrite-each-other.md): Finds updates that overwrite newer data because two things wrote at once.
- [Half-finished failures](https://atlasscope.dev/prompts/half-finished-failures.md): Finds what is left behind when something fails halfway.
- [Lost place: Back, refresh and shared links](https://atlasscope.dev/prompts/lost-place-back-refresh-and-shared-links.md): Finds the everyday friction where Back closes the wrong thing, a filter resets, or a pasted link opens the wrong view.
- [Unit tests that give false comfort](https://atlasscope.dev/prompts/unit-tests-that-give-false-comfort.md): Finds features whose automated checks all pass and that still break for real people.
- [Tests that can't fail](https://atlasscope.dev/prompts/tests-that-cant-fail.md): Finds tests that pass no matter what.
- [Outside input reaching sensitive actions](https://atlasscope.dev/prompts/outside-input-reaching-sensitive-actions.md): Follows input from outsiders to the places it can do harm.
- [Risk from outside packages](https://atlasscope.dev/prompts/risk-from-outside-packages.md): Checks what a harmful outside package or pull request could reach.
- [Security checks that fail open](https://atlasscope.dev/prompts/security-checks-that-fail-open.md): Finds security checks that let people in when something goes wrong.
- [What the tests really prove](https://atlasscope.dev/prompts/what-the-tests-really-prove.md): Asks what each test really proves, then adds one check that would catch a real break.
- [Settings that differ in production](https://atlasscope.dev/prompts/settings-that-differ-in-production.md): Finds settings that behave differently in production than anyone expects.
- [Why production might not match main](https://atlasscope.dev/prompts/why-production-might-not-match-main.md): Lists every way 'merged' can fail to mean 'live'.
- [Check time that catches nothing](https://atlasscope.dev/prompts/check-time-that-catches-nothing.md): Finds paid automated-check minutes spent on checks that never catch anything.
- [Background jobs that hang or stop halfway](https://atlasscope.dev/prompts/background-jobs-that-hang-or-stop-halfway.md): Finds scheduled jobs and background workers that can hang, overlap or stop halfway.
- [Packages nobody uses, or that work by luck](https://atlasscope.dev/prompts/packages-nobody-uses-or-that-work-by-luck.md): Finds packages nobody uses and packages that work only by accident.
- [Missing vs zero vs stale](https://atlasscope.dev/prompts/missing-vs-zero-vs-stale.md): Finds where 'we don't know' turns into 0 and then into a wrong total.
- [Where did this number come from?](https://atlasscope.dev/prompts/where-did-this-number-come-from.md): Checks that every number on screen can be traced back to a source document.
- [Old records, new meanings](https://atlasscope.dev/prompts/old-records-new-meanings.md): Finds old records that read differently once today's rules are applied to them.
- [Imports that fail quietly](https://atlasscope.dev/prompts/imports-that-fail-quietly.md): Finds rows that go missing or land on the wrong record during import, with no error.
- [Links between records worth adding](https://atlasscope.dev/prompts/links-between-records-worth-adding.md): Finds links between records worth adding, and ones the evidence does not support.
- [Code health, with proof](https://atlasscope.dev/prompts/code-health-with-proof.md): A general health check of the code that refuses findings without evidence.
- [Out-of-date agent instructions](https://atlasscope.dev/prompts/out-of-date-agent-instructions.md): Finds instructions that send agents to files and patterns that no longer exist.
- [Database changes during a deploy](https://atlasscope.dev/prompts/database-changes-during-a-deploy.md): Checks that each database change works with the old code that is still live while a release finishes.

## Skills

- [animate](https://atlasscope.dev/skills/animate.md) by Emil Kowalski
- [animate-expo](https://atlasscope.dev/skills/animate-expo.md) by Emil Kowalski
- [animation-performance](https://atlasscope.dev/skills/animation-performance.md) by Emil Kowalski
- [animation-vocabulary](https://atlasscope.dev/skills/animation-vocabulary.md) by Emil Kowalski
- [apple-design](https://atlasscope.dev/skills/apple-design.md) by Emil Kowalski
- [ask-sonner](https://atlasscope.dev/skills/ask-sonner.md) by Emil Kowalski
- [break-ui](https://atlasscope.dev/skills/break-ui.md) by Emil Kowalski
- [color-expert](https://atlasscope.dev/skills/color-expert.md) by meodai
- [copywriting](https://atlasscope.dev/skills/copywriting.md) by Corey Haines
- [cro](https://atlasscope.dev/skills/cro.md) by Corey Haines
- [design](https://atlasscope.dev/skills/design.md)
- [design-everyday-things](https://atlasscope.dev/skills/design-everyday-things.md) by Wondel.ai
- [emil-design-eng](https://atlasscope.dev/skills/emil-design-eng.md) by Emil Kowalski
- [find-animation-opportunities](https://atlasscope.dev/skills/find-animation-opportunities.md) by Emil Kowalski
- [fix-writing](https://atlasscope.dev/skills/fix-writing.md)
- [hallmark](https://atlasscope.dev/skills/hallmark.md) by Hallmark contributors
- [hooked-ux](https://atlasscope.dev/skills/hooked-ux.md) by Wondel.ai
- [improve-animations](https://atlasscope.dev/skills/improve-animations.md) by Emil Kowalski
- [improve-retention](https://atlasscope.dev/skills/improve-retention.md) by Wondel.ai
- [ios-hig-design](https://atlasscope.dev/skills/ios-hig-design.md) by Wondel.ai
- [lean-ux](https://atlasscope.dev/skills/lean-ux.md) by Wondel.ai
- [microinteractions](https://atlasscope.dev/skills/microinteractions.md) by Wondel.ai
- [mobile-native](https://atlasscope.dev/skills/mobile-native.md) by Emil Kowalski
- [oklch-skill](https://atlasscope.dev/skills/oklch-skill.md) by Jakub Krehel
- [pick-ui-library](https://atlasscope.dev/skills/pick-ui-library.md) by Emil Kowalski
- [product-marketing](https://atlasscope.dev/skills/product-marketing.md) by Corey Haines
- [prototype](https://atlasscope.dev/skills/prototype.md) by Emil Kowalski
- [refactoring-ui](https://atlasscope.dev/skills/refactoring-ui.md) by Wondel.ai
- [review-animations](https://atlasscope.dev/skills/review-animations.md) by Emil Kowalski
- [steve-jobs-design-review](https://atlasscope.dev/skills/steve-jobs-design-review.md) by Wondel.ai
- [stop-slop](https://atlasscope.dev/skills/stop-slop.md) by Hardik Pandya
- [swiftui-pro](https://atlasscope.dev/skills/swiftui-pro.md) by Paul Hudson
- [top-design](https://atlasscope.dev/skills/top-design.md) by Wondel.ai
- [ux-heuristics](https://atlasscope.dev/skills/ux-heuristics.md) by Wondel.ai
- [ux-writing](https://atlasscope.dev/skills/ux-writing.md) by Christopher Greer
- [web-typography](https://atlasscope.dev/skills/web-typography.md) by Wondel.ai
- [write-swift](https://atlasscope.dev/skills/write-swift.md) by Emil Kowalski
- [codebase-design](https://atlasscope.dev/skills/codebase-design.md) by Matt Pocock
- [database](https://atlasscope.dev/skills/database.md)
- [diagnosing-bugs](https://atlasscope.dev/skills/diagnosing-bugs.md) by Matt Pocock
- [domain-modeling](https://atlasscope.dev/skills/domain-modeling.md) by Matt Pocock
- [engineer](https://atlasscope.dev/skills/engineer.md)
- [grill-me](https://atlasscope.dev/skills/grill-me.md) by Matt Pocock
- [grill-with-docs](https://atlasscope.dev/skills/grill-with-docs.md) by Matt Pocock
- [grilling](https://atlasscope.dev/skills/grilling.md) by Matt Pocock
- [improve-codebase-architecture](https://atlasscope.dev/skills/improve-codebase-architecture.md) by Matt Pocock
- [research](https://atlasscope.dev/skills/research.md) by Matt Pocock
- [retro](https://atlasscope.dev/skills/retro.md) by Matt Pocock
- [tdd](https://atlasscope.dev/skills/tdd.md) by Matt Pocock
- [to-questionnaire](https://atlasscope.dev/skills/to-questionnaire.md) by Matt Pocock
- [wait-what](https://atlasscope.dev/skills/wait-what.md) by Matt Pocock
- [wizard](https://atlasscope.dev/skills/wizard.md) by Matt Pocock
- [writing-for-agents](https://atlasscope.dev/skills/writing-for-agents.md) by Matt Pocock
- [diagnose](https://atlasscope.dev/skills/diagnose.md)
- [rethink](https://atlasscope.dev/skills/rethink.md)
- [verify](https://atlasscope.dev/skills/verify.md)
- [ceo](https://atlasscope.dev/skills/ceo.md)
- [harsh-review](https://atlasscope.dev/skills/harsh-review.md)
- [recommend](https://atlasscope.dev/skills/recommend.md)
- [viewport-proof](https://atlasscope.dev/skills/viewport-proof.md)

## Optional

- [Source on GitHub](https://github.com/AtlasScope/agent-kit)
- [Credits](https://github.com/AtlasScope/agent-kit/blob/main/CREDITS.md)