user-scenarios.md

User-scenario matrix

One table that maps the real user journeys to what covers them today and what the Nightly rewrite should cover next (#358). It is the map; the tests are elsewhere. The journeys below are the draft from the issue, kept as the proposal; corrections belong in review, not in a quiet edit.

Two things this page deliberately does not decide:

• Nightly target is the owner's call. Every cell reads TBD (owner decides), with the proposal from #358 in parentheses where the issue made one, so the column can be signed off in one pass instead of reading a decision this PR invented.
• Which journeys are in scope at all is still the proposal in the issue body. A row is added or dropped by that discussion, not here.

Rules for the Today column: it must point at something concrete — a test file, .github/workflows/nightly.yml, .github/workflows/live-e2e.yml, or core/dev/eval/specs.mjs — or say none. A row may not claim coverage without a pointer.

ID	Scenario	Persona	Surfaces	Plugins	Today	Nightly target	Blocked by
P1	Cold start → first message sendable	personal solo	local Desktop, local web	core	none	TBD (owner decides); proposal: perf, #357	#356, #357
P2	Single Bot chat with one tool call	personal solo	local web	core	.github/workflows/live-e2e.yml, core/test/rounds.test.js	TBD (owner decides); proposal: live E2E + E2E mock	—
P3	Tree run: Chief → children → one combined answer	personal solo	local web	core, tree	core/dev/eval/specs.mjs (P3-tree-split), core/test/bot-tree.test.js	TBD (owner decides); proposal: live E2E	—
P4	Invite a friend, message their Bot, approve the held reply	personal + friend	local web	social	core/dev/eval/specs.mjs (P4-invite-approve), core/test/held-reply.test.js	TBD (owner decides); proposal: live E2E	—
P5	Face-to-face room: join, @Bot, leave	personal + friend	local web	social (face rooms)	core/test/face-people.test.js	TBD (owner decides)	—
P6	Timeline: read swimlanes after a multi-Bot run	personal solo	local web	timeline	core/dev/eval/specs.mjs (P6-timeline)	TBD (owner decides); proposal: live E2E + perf	—
P7	Long session: 500+ messages, scroll and reopen	personal solo	local web	core	none	TBD (owner decides); proposal: perf, #356/#357	#356, #357
P8	Background tab hidden 5 min, then return	personal solo	local web	core	none	TBD (owner decides)	#259
X1	Slow or failing model API: waiting, timeout and retry are clear	any	local web	core	none	TBD (owner decides); proposal: E2E mock	—
X2	Permission denied / no grant	team member	deployed instance (#352)	social, grants	core/test/authz.test.js	TBD (owner decides)	#51, #352
T1	Assign a task, the owner sees Needs you, approves	team	deployed instance (#352)	inbox, social	none	TBD (owner decides)	#349, #149, #352
T2	Approval latency across two instances	team	deployed instance (#352)	social	none	TBD (owner decides)	#352
T3	Reminder: an approval waits too long	personal, team	local web	reminders (#353)	none	TBD (owner decides)	#353
T4	Scheduled morning summary arrives	personal	local web	reminders (#354)	none	TBD (owner decides)	#354
T5	Reminder while the desktop app is closed (24/7)	personal	external channel (#355)	reminders, #352	none	TBD (owner decides)	#355, #352

How to use it

1. A journey changes → the row changes the same day. The ids are stable: P* are personal-mode journeys, X* cross-cutting failures, T* team-mode journeys.
2. core/dev/eval/specs.mjs follows these ids. Its journey field is the P-id of the row it scores (P3, P4, P6 today), so a new P-row that gets a canonical rubric needs a spec with the same id — and the id in this table is what run.mjs reports.
3. The Nightly rewrite reads the Nightly target column once the owner signs it (#357 wires the metrics into it). Until then, nightly.yml stays as it is.

Out of scope here

Rewriting .github/workflows/nightly.yml and implementing any missing test. This page is signed first; the suites follow.