[mvp] MVP definition, personas & UX journeys — workstream kickoff

[lalalune]

This is the coordination thread for the MVP definition, personas & UX journeys workstream of the LifeOps Personal Assistant MVP (GitHub project 15).

What the MVP is

Chat, onboarding, the current views, and centrally LifeOps — all scheduling, calendar, coordination, reminders, goals, todos, tasks, personal assistant. It serves business users, children, adults with ADHD/ADD/Asperger's/autism, neurotypical people, and elderly people — no therapy language, no special rails. One interface; persona differences are owner facts + structural scheduler knobs, which is how the runtime is already built.

Guiding constraint: minimize additional scope. We are turning what exists into an MVP by fixing, testing, verifying and validating. Prefer deleting/simplifying over adding. Every piece of work in this workstream is scenario verification or scenario authoring against existing features — no new features, no new packs, no persona modes.

Doctrine adopted (do not relitigate here)

• Chat is the primary surface; views are information surfaces the agent (or the user via chat) drives. Views: uniform top bar, no side panels, no suggestion chips, agent-proactive-on-view-switch, view-scoped actions.
• One scheduler: everything scheduled routes through plugins/plugin-scheduling/src/scheduled-task/runner.ts; the runner matches structural fields only, never promptInstructions.
• Orange accent only, no blue, minimal UI, no card chrome.
• Fail fast, no fallback sludge; logger not console; strong types.
• Persona scenarios follow the authoring contract at plugins/plugin-personal-assistant/test/scenarios/_catalogs/LIFEOPS_PERSONA_SCENARIO_AUTHORING.md (tier rubric T1-T4, no echo-satisfiable assertions, outcomes not routing, two-phase commit, tone judges that fail guilt framing and therapy roleplay).

Research doc

Full analysis — current state with file references, the 7 personas and their journeys, the real-life scenario matrix, open questions with answers:

packages/docs/ongoing-development/research/01-mvp-product-personas.md

Headline numbers: 271 live-only scenarios exist; the 8 persona packs are 212 authored / 134 verified. The gaps are concentrated exactly where the MVP promise lives: ADHD capture 8/28 verified, ADHD follow-through 5/24, shift-rotation 4/22, plus zero coverage for a child as the user, a student's report-deadline, and the onboarding journey itself.

Issues (most important first)

1. [mvp] Verify ADHD capture + follow-through packs (A1 8/28, A2 5/24) with live-LLM runs — P0
2. [mvp] Verify shift-rotation pack B2 (4/22) — irregular-sleep re-anchoring — P0
3. [mvp] Author + verify child-voice and student report-deadline scenarios (zero coverage today) — P0
4. [mvp] Author + verify first-run onboarding journey scenarios (fast-start + customize → first reminder) — P1
5. [mvp] Verify elderly week-1 loop: daily check-in, missed-checkin follow-up, quiet-user watcher — P1
6. [mvp] Author + verify non-sleep life-goal grounding scenarios (goals corpus is sleep-only) — P1
7. [mvp] Verify fresh-user MVP view set + finish F1 neurotypical-control pack (24/32) — P2
8. [mvp] Fix dead docs/automation-glossary.md link in root CLAUDE.md/AGENTS.md — P2

How to claim work

1. Comment on the issue you are taking ("claiming — session/branch name").
2. Move the card on project 15 to In Progress.
3. Branch as fix|test|docs/<issue#>-<slug> off develop; one PR per issue against develop. Sync (git fetch origin && git rebase origin/develop) before opening or updating.
4. Before claiming, check the issue isn't already closed and git log origin/develop on the target files — the swarm merges fast.

Evidence rules (changed mechanics — read this)

Evidence is now posted inline in the GitHub issue/PR, not committed to .github/issue-evidence/:

• MP4 for video walkthroughs (renders inline in GitHub). Always build/deploy the current tree before capturing.
• JPG for screenshots (size). For bug fixes, include the wrong behavior before the fix, then the fixed behavior. Desktop + mobile where a UI surface is touched.
• Live-LLM trajectories: run packages/scenario-runner/bin/eliza-scenarios run <scenario> --report <out> against a live model (not the proxy, not a mock), then read the trajectory by hand and paste annotated excerpts inline. A captured-but-unread trajectory is not evidence.
• Logs: structured [ClassName] backend logs showing the real code path firing; frontend console/network where relevant.
• Domain artifacts: the scheduled-task rows, goal records, state-log transitions your run created — queried and pasted.
• Scenario verification lands as a catalog status: "verified" flip with a model + judge-score note (existing precedent in the catalog files), and node packages/scripts/check-lifeops-persona-catalog-coverage.mjs --json before/after output in the PR.
• Anything that doesn't apply: write N/A — <reason> explicitly, never leave it blank.

Out of scope for this workstream

New default packs, child/elderly UI variants, onboarding rewording, visual-timeline/gamification features, CI lanes for the live-only corpus, new connectors, launcher redesigns. If a scenario run exposes a real behavior defect that's non-trivial to fix, file a linked issue with the failing trajectory attached rather than growing your PR.

Filed issues

• [mvp] Verify ADHD capture + follow-through packs (A1 8/28, A2 5/24) with live-LLM runs #14350 — P0 [mvp] Verify ADHD capture + follow-through packs (A1 8/28, A2 5/24) with live-LLM runs
• [mvp] Verify shift-rotation pack B2 (4/22) — irregular-sleep re-anchoring #14351 — P0 [mvp] Verify shift-rotation pack B2 (4/22) — irregular-sleep re-anchoring
• [mvp] Author + verify child-voice and student report-deadline scenarios (zero coverage today) #14352 — P0 [mvp] Author + verify child-voice and student report-deadline scenarios (zero coverage today)
• [mvp] Author + verify first-run onboarding journey scenarios (fast-start + customize → first reminder) #14353 — P1 [mvp] Author + verify first-run onboarding journey scenarios (fast-start + customize → first reminder)
• [mvp] Verify elderly week-1 loop: daily check-in, missed-checkin follow-up, quiet-user watcher #14354 — P1 [mvp] Verify elderly week-1 loop: daily check-in, missed-checkin follow-up, quiet-user watcher
• [mvp] Author + verify non-sleep life-goal grounding scenarios (goals corpus is sleep-only) #14355 — P1 [mvp] Author + verify non-sleep life-goal grounding scenarios (goals corpus is sleep-only)
• [mvp] Verify fresh-user MVP view set + finish F1 neurotypical-control pack (24/32) #14356 — P2 [mvp] Verify fresh-user MVP view set + finish F1 neurotypical-control pack (24/32)

---

Design doc · MVP board