swarm-ui-reviewer
Skill2 starsChanged 19 days agoApache-2.0View on GitHub
Extra rules for Swarm ui_reviewer agents (role ui_reviewer) reviewing UI work. Use together with the `swarm` skill when your kickoff names this skill or your assignment has a `## Workflow` step naming you "ui_reviewer".
In this repository26 other files it gives its agents.
- AGENTS.mdAGENTS.md
- skills/vendor/vercel-react-native-skills/AGENTS.mdAGENTS.md
- swarm-advisorSkill
- swarm-batchingSkill
- swarm-coderSkill
- swarm-debuggerSkill
- swarm-designerSkill
- swarm-mechanicalSkill
- swarm-orchestratorSkill
- swarm-researcherSkill
- swarm-reviewerSkill
- swarmSkill
- swarm-spikeSkill
- swarm-workflowsSkill
- building-componentsSkill
- expo-design-systemSkill
- expo-native-uiSkill
- graphifySkill
- mobile-android-designSkill
- mobile-ios-designSkill
- ponytail-debtSkill
- ponytail-reviewSkill
- ponytailSkill
- ui-ux-pro-maxSkill
- vercel-react-native-skillsSkill
- web-design-guidelinesSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 0 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 14
Visual and interaction quality: run web-design-guidelines against every changed UI file (web) — layout, spacing, typography, color usage, …
- Very specific: line 15
Component API quality: check new or changed components against building-components — prop surface, composition over boolean-prop …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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
Reviewing UI with Swarm
Follow the swarm skill first; this adds to it. Then read the swarm-reviewer skill in full — it is not in your kickoff, but everything in it applies to you: review order (spec/acceptance → tests → correctness → security → simplicity/ponytail-review → conventions), walking the package unit by unit, never editing files, and the verdict/findings contract below. This skill only adds the UI-specific checks and sources.
Verdict and findings, restated
completed carries verdict (pass | changes_requested | blocked) and findings[] ({severity: critical|major|minor|nit, file, line, summary, unit}), same as swarm-reviewer. pass is allowed only when every remaining finding is nit/minor; a missing unit is always at least major. Never edit files.
UI-specific checks, after the standard review order
- Visual and interaction quality: run
web-design-guidelinesagainst every changed UI file (web) — layout, spacing, typography, color usage, states (empty/loading/error/success), responsiveness. - Component API quality: check new or changed components against
building-components— prop surface, composition over boolean-prop proliferation, sensible defaults, reusability. - Design artifact honoured: if the task (or one it depends on) has a
designartifact registered, read it (swarm_read) and confirm the built UI matches it — screens, states, tokens, interaction. A deviation that isn't called out in the coder'scompletedsummary is a finding. - Mobile: for React Native / Expo or native iOS/Android surfaces, apply whichever of
mobile-ios-design,mobile-android-design,expo-native-ui,vercel-react-native-skillsfits the platform, plus theui-ux-pro-maxpre-delivery checklist (references/pro-rules.md): safe areas, touch targets ≥ 44pt (iOS) / 48dp (Android), dynamic type / font scaling, reduced motion, dark mode, contrast.
Cite the rule source in every UI finding (e.g. "web-design-guidelines: insufficient contrast" or "ui-ux-pro-max pro-rules.md: touch target 32pt, needs ≥44pt") so the coder knows which skill to re-read.
Store the screenshots you capture under ~/.swarm/screenshots/<ITEM-KEY>/ (your own item key), never ~/.swarm/attachments, which is the daemon's swept request-attachment store.
Headless Chrome: keep it in your dirs, close it before you finish
Chrome copies its app bundle into com.google.Chrome.code_sign_clone at every launch and deletes the copy only on a clean quit, so a Chrome left running leaks hundreds of thousands of inodes. If you start Chrome for screenshots or inspection:
- Launch it with
--user-data-dirpointing inside your own scratchpad or work dir, never a shared or system path. - Close it before your
completedorhandoffcheckpoint, and on any early exit: sendBrowser.closeover CDP, or SIGTERM it and wait for it to exit (never SIGKILL first).
Your review tree is a fresh checkout. Build its graph first (rule 9b), then use graphify affected on each changed symbol to find callers the diff may have broken.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
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.

