UX Flow Auditor

UX Flow Auditor

Task-oriented UX — how many clicks to achieve a goal, where users get stuck.

All artifacts go under .heyeddi/docs/ and .heyeddi/audits/ — never repo root.

When to use

• "How many clicks to update settings?"
• Post-ship ease-of-use check
• Compare desktop vs mobile flow (separate flow JSON or viewport)
• Product acceptance: task completes within click budget

Subagents (default)

See reference/subagents.md. Run trace_flow.py via Task shell (needs dev server + Playwright).

Pipeline

init_ux_flows          → ux-flows.md index + example .flow.json
define steps in ux-flows/<task>.flow.json
trace_flow --task-id   → ux-flows/<task>.md + audit screenshot
agent adds friction notes to report

Instructions

1. Ensure .heyeddi/product.md defines user tasks (optional context).
2. python scripts/init_ux_flows.py --project-root <root> once per project.
3. Add or edit .heyeddi/docs/ux-flows/<task-id>.flow.json with steps (expect, fill, click).
4. Start dev server; set DEV_SERVER_URL if not http://localhost:5173.
5. python scripts/trace_flow.py --task-id <id> --project-root <root> --check
6. Append Friction notes to the generated markdown report.

Flow JSON schema

{
  "task_id": "update-profile",
  "goal": "User saves profile",
  "start_route": "/settings",
  "max_clicks": 4,
  "viewport_width": 1440,
  "steps": [
    { "action": "expect", "selector": ".settings .p-inputtext", "label": "Form visible" },
    { "action": "fill", "selector": ".settings input >> nth=0", "value": "Alex", "label": "Name" },
    { "action": "click", "selector": "button:has-text('Save')", "label": "Save" }
  ]
}

.heyeddi/ outputs

Chain

• After @heyeddi-handoff / @heyeddi-design — verify tasks are completable
• @visual-auditor — layout proof; this skill — interaction proof
• @engineering-excellence — separate concern (code structure)