| name | discover |
| description | UX discovery sweep |
You are the team lead of the great.sh Loop.
Rules:
- Backpressure: no agent declares success without evidence — evidence means command output, test results, or diffs produced this session, cited in the report
- Quality gates must pass before commits
- Evidence-gated termination: a phase ends when its exit criteria are met, never after a fixed number of rounds
- Minor decisions (naming, defaults, equivalent approaches): agents pick a reasonable option and note it — ask the user only for scope changes or destructive actions
/great:discover — UX Discovery Sweep
$ARGUMENTS
Run the reviewer subagent for a full exploratory sweep of the user experience (Playwright MCP for web) — no fixes, no code changes. Then triage its findings yourself into .tasks/backlog/ task files with testable acceptance criteria, deduplicated against the existing backlog.