# Feedback Lab triage preview Purpose: review category, priority, factual summary and next steps beside a report. All five shared fixtures are synthetic. Suggestions are canned, not provider output. There is no backend, network API, login, private key or persisted review decision. Accept/dismiss affects only local demo/hash state. Original status stays New. Select a report and response. Analyze → queued → Advance simulation → running → Advance simulation → completed/error/invalid. Completed permits accept/dismiss. Error/invalid permits retry. Reset returns the first report, success and idle. For recovery, choose Valid suggestion while in error/invalid, then Retry analysis. Actions move focus to the status announcement so hidden actions cannot strand focus. Each state also has a shortcut link. Refresh preserves the hash selection: `#report=login-button&state=completed&outcome=success`. States: `idle`, `queued`, `running`, `completed`, `error`, `invalid`, `accepted`, `dismissed`. Reports: `login-button`, `csv-export`, `status-question`, `insufficient-context`, `prompt-injection`. Invalid IDs/states fall back to the first report/idle. Progress is manual for repeatability. Review at 1440px and 390px. Use Tab/Enter to analyze, advance twice, then accept. Check long injection text, focus visibility, announced progress, responsive columns, retry and refresh. Busy disables duplicate submission/report changes. Questions: Is manual progression clear? Can reviewers distinguish acceptance from a status update? Are failure and invalid output distinct? Is keyboard focus retained? Exercise 4.04 handoff: connect this UI to Lite's built-in `ctx.agent` and the provider key configured in the project's AI page. Add authenticated ownership checks, asynchronous session/run persistence and event polling, exact JSON output validation, busy/quota/uncertain-submit handling, session cleanup and review storage. Treat report text as untrusted data. Never execute next steps or change report status on accept. No live agent adapter is included in exercise 4.03.