designing-interaction-states
Skill111 starsChanged 18 days agoMITView on GitHub
Designs the states screens actually spend their time in — first-run and empty states, loading and skeletons, inline validation and error messages, success feedback, destructive confirmations, and the microcopy for each — so a product feels finished instead of only working on the happy path. Use when building a form, list, dashboard, upload, checkout or settings screen, adding an async action, writing error or empty-state copy, or reviewing a UI that shows blank pages, raw errors or spinners forever.
In this repository41 other files it gives its agents.
- building-accessible-interfacesSkill
- wiring-application-configSkill
- shipping-build-artifactsSkill
- permissioning-ci-workflowsSkill
- concurrent-branchesSkill
- shipping-across-surfacesSkill
- writing-defect-reportsSkill
- reporting-derived-metricsSkill
- guarding-destructive-operationsSkill
- running-github-actions-efficientlySkill
- keeping-git-repos-cleanSkill
- designing-marketing-pagesSkill
- rendering-untrusted-contentSkill
- reproducing-ci-locallySkill
- running-resumable-sync-jobsSkill
- normalizing-text-for-measurementSkill
- calibrating-thresholds-and-baselinesSkill
- designing-interfacesSkill
- verifying-external-behaviorSkill
- building-go-projectsSkill
- building-browser-extensionsSkill
- designing-python-apisSkill
- building-python-clisSkill
- improving-python-code-qualitySkill
- building-python-communitiesSkill
- documenting-python-librariesSkill
- deferring-heavy-dependenciesSkill
- reviewing-python-librariesSkill
- building-llm-backed-featuresSkill
- building-python-mcp-serversSkill
- packaging-python-librariesSkill
- optimizing-python-performanceSkill
- setting-up-python-librariesSkill
- managing-python-releasesSkill
- auditing-python-securitySkill
- testing-python-librariesSkill
- building-python-web-appsSkill
- building-rust-cratesSkill
- building-scala-projectsSkill
- building-swift-appsSkill
- shipping-swift-appsSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 4 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 60
Never an indefinite spinner: every request has a timeout that becomes an error state with a retry.
- Very specific: line 75
Use the right input types (email, tel, number, date, autocomplete attributes) so phones show the right keyboard and browsers can autofill.
- Distinctive: line 91
Never show stack traces, raw exception text or status codes to end users.
- Distinctive: line 108
Prefer undo over confirmation for reversible actions ("Contact archived.
- Distinctive: line 122
No blame ("You entered an invalid…" → "Enter a valid…"), no jokes in errors.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Designing Interaction States
Most screens are designed in one state: full of realistic data, with every request succeeding. Users mostly see the others — empty on day one, loading on a slow network, an error on a typo. Those states are where an app feels either finished or neglected.
Contents
- The state inventory
- Empty and first-run states
- Loading
- Forms and validation
- Errors
- Success and feedback
- Destructive actions
- Microcopy rules
- Checklist
The state inventory
For every screen, list and design each of these before calling it done:
| State | Question it answers |
| First run / empty | "What is this and what do I do first?" |
| Loading | "Is it working?" |
| Partial / some data | "Does it still look intentional with one item?" |
| Full / overflow | "What happens with 500 items or a very long name?" |
| Error | "What went wrong, and what do I do now?" |
| Success | "Did it work?" |
| Offline / permission denied | "Why can't I do this?" |
Seed local data for each and screenshot them; reviewers should see every state.
Empty and first-run states
An empty state is the product's best onboarding surface. Replace "No items." with:
- a one-line explanation of what will appear here,
- the single primary action that fills it ("Import contacts", "Create your first project"), as a real button,
- optionally a secondary path (sample data, docs link).
Distinguish empty because new from empty because filtered ("No results for 'acme' — clear filters") from empty because of an error (that is an error state, not an empty one).
Loading
- Under ~300ms: show nothing (a flashing spinner reads as jank).
- Content areas: skeletons shaped like the content, so the layout does not jump.
- Actions: disable the button and show progress in it ("Saving…"), preventing double submits.
- Over ~10s or multi-step: real progress (steps or percent) and the option to leave.
- Never an indefinite spinner: every request has a timeout that becomes an error state with a retry.
Forms and validation
- Labels above fields, visible always; help text below the label, not in the placeholder.
- Mark optional fields ("(optional)") rather than starring required ones, when most are required.
- Validate on blur or on submit, not on every keystroke; never show an error before the user has had a chance to type.
- On submit with errors: keep everything the user typed, show a summary at the top linking to each field, move focus to the first invalid field.
- Error text sits next to its field, says what is wrong and how to fix it, and is
linked with
aria-describedby(see thebuilding-accessible-interfacesskill). - Use the right input types (
email,tel,number,date,autocompleteattributes) so phones show the right keyboard and browsers can autofill. - One primary button per form, at the end, labelled with the outcome ("Create account", not "Submit").
Errors
A good error message has three parts: what happened, why (if known), what to do next.
| Bad | Good |
| "Error 500" | "We couldn't save your changes. Your edits are still here — try again in a moment." |
| "Invalid input" | "Enter a date in the future." |
| "Failed to fetch" | "You're offline. We'll retry when you reconnect." |
- Never show stack traces, raw exception text or status codes to end users; log them with a reference ID and show the ID.
- Errors are recoverable where possible: a retry button, preserved input, a link to the place to fix the cause (e.g. billing settings).
- Page-level failures get a designed page (404, 500, expired link) with navigation back into the app, not the framework default.
Success and feedback
- Confirm every action that changes data: inline ("Saved ✓" next to the button) for small edits, a toast for background actions, a full success screen for milestones (account created, payment complete) that says what happens next.
- Toasts auto-dismiss (~5s) except for errors and anything with an action ("Undo").
- Optimistic UI is fine for low-risk actions if failures roll back visibly.
Destructive actions
- Prefer undo over confirmation for reversible actions ("Contact archived. Undo").
- For irreversible ones, confirm with a dialog that names the object and the consequence ("Delete 'Q3 report'? This removes it for everyone and can't be undone.") and a button labelled with the action ("Delete report"), styled as danger — never a generic "OK".
- For high-stakes deletes (an account, a workspace), require typing the name.
Microcopy rules
- Plain words, sentence case, second person ("Your invoices"), active voice.
- Buttons are verbs naming the outcome; links say where they go.
- Be specific with numbers and names ("3 contacts imported, 2 skipped — see why").
- Same term for the same thing everywhere (pick "workspace" or "team", not both).
- No blame ("You entered an invalid…" → "Enter a valid…"), no jokes in errors.
Checklist
- Every screen has designed empty, loading, error and success states
- No spinner without a timeout and an error path
- Forms keep input on error, focus the first invalid field, and use real labels
- Error messages say what happened and what to do; no raw exceptions shown
- Destructive actions use undo or a specific, danger-styled confirmation
- Screenshots of each state attached to the PR
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
Reviews can't be read right now.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

