interaction-reviewer — Claude Code skill: what it does
interaction-reviewer is a Claude Code skill that reviews interaction flow and control behavior. It detects missing feedback, unnecessary steps, and unsafe UI state transitions.
MIT ★ 529 updated 2026-11-17
npx skills add ShuaiMXu/happyhands --skill interaction-reviewer -a claude-codeWhat interaction-reviewer does
The skill analyzes end-to-end user flows to uncover unnecessary steps, missing feedback, and unsafe interface transitions. It examines loading, success, error, and empty states across product tasks. During the evaluation, deliberate friction for safety is separated from usability flaws.
Static screenshots are insufficient for interaction claims. The workflow requires testing controls directly with keyboard and pointer inputs, tracking focus shifts, feedback, and error recovery. When evidence lacks validation or defects remain unproven, the candidate stays unadmitted and no edits are applied.
What it helps with
- detect unsafe UI state transitions
- identify missing feedback during user actions
- separate intentional safety controls from usability friction
- audit keyboard focus and pointer recovery steps
Example prompts for the agent
- Review the checkout flow for extra steps and missing feedback.
- Audit state transitions between loading and error in the signup form.
- Inspect control behavior and focus states during keyboard navigation.
How to install interaction-reviewer
1Skills CLI (Vercel)
npx skills add ShuaiMXu/happyhands --skill interaction-reviewer -a claude-code2Manually: put the files in place
~/.claude/skills/interaction-reviewer/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/interaction-reviewer/
1Skills CLI (Vercel)
npx skills add ShuaiMXu/happyhands --skill interaction-reviewer -a codex2Manually: put the files in place
.agents/skills/interaction-reviewer/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add ShuaiMXu/happyhands --skill interaction-reviewer -a cursor2Manually: put the files in place
.cursor/skills/interaction-reviewer/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install ShuaiMXu/happyhands skills/product-design/interaction-reviewer2Manually: put the files in place
.github/skills/interaction-reviewer/SKILL.md1Skills CLI (Vercel)
npx skills add ShuaiMXu/happyhands --skill interaction-reviewer -a gemini-cli2Platform CLI
gemini skills install https://github.com/ShuaiMXu/happyhands --path skills/product-design/interaction-reviewerManually: put the files in place
.devin/skills/interaction-reviewer/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about interaction-reviewer
How to install interaction-reviewer in Claude Code?
Run npx skills add ShuaiMXu/happyhands --skill interaction-reviewer -a claude-code or save the file to ~/.claude/skills/interaction-reviewer/SKILL.md.
What does interaction-reviewer do?
It reviews task flows, tracks transitions across loading, error, and success states, and identifies missing interface feedback.
What evidence does interaction-reviewer require?
The skill requires documented task steps, full state definitions, and control testing via keyboard or pointer. Static screenshots alone are insufficient for interaction claims.
Source text
SKILL.md · MIT
--- name: interaction-reviewer description: "Review interaction flow, feedback, control behavior and state transitions for a critical product task." --- # Interaction Reviewer Find unnecessary steps, missing feedback and unsafe state transitions. ## Required evidence - task steps - interactive states - loading, success, error and empty states ## Decisions - Follow the task end to end - Distinguish friction from safety controls - Describe expected state transitions ## Boundary Static screenshots are insufficient for interaction claims. Return structured evidence, findings, proposed updates and limitations. Use `needs-evidence` when required inputs or professional review are missing. Never invent an Experience Score or promote a project observation into a universal rule. ## Execution contract and admission The bounded handler supplies: Unnamed-control candidates from reached capture DOM. The professional method still requires: Replay the specific control with keyboard and pointer; observe focus, feedback and error/recovery. Source words or control counts do not establish behavior. Deliver into the current Run's existing experience audit or decision; do not create a competing report. Cite the frozen Skill path and exact rule, original state and source, observed failure, task consequence and the reasoning connecting them. Record preservation/counterevidence. With no demonstrated defect, keep the candidate unadmitted and make no change. `complete` describes this handler's bounded output, not professional review or human acceptance.
© ShuaiMXu · MIT · ShuaiMXu/happyhands