emrg: GUI session header + session-list name refresh + result-list pane fix

[argszero]

Summary

Implements two host rants:

1. rant 2026-08-13T20:41:37 — per-session sticky header (project/name(id) / project/id) + rename refresh bug fix for the open-sessions sidebar.
2. rant 2026-08-13T20:49:45 — the artifact pane (#result-list) was always visible because .result-list { display:flex } (same specificity, defined later) overrode .result-pane { display:none }, so the file/artifact tabs were not independent.

Changes

• app.js: add .session-header child per .session-view; add sessionProjectName() + renderSessionHeader(); call on activate + refreshSessions (which now re-renders open-sessions list too).
• chat.js: clear() preserves .session-header.
• layout.css: add .session-header style; remove redundant display:flex / flex-direction:column from .result-list (so pane visibility is controlled solely by .result-pane/.result-pane.active).
• renderer.smoke.test.js: add msgCount() helper (excludes header) + update 3 P3-s2 assertions; add a CSS source-level regression test asserting .result-list does not redeclare display/ flex-direction.
• Agent.md: GUI test count 237 → 238 (renderer smoke 115 → 116).

Verification

• node --check on touched JS: OK
• cd emrg/gui && npm test: 238/238 pass
• uv run pytest tests/ -q: 801 passed
• uv run python -c "from emrg.client.app import run_client" + uv run python -m emrg --help: OK

[argszero]

❌ Needs fix: unrelated .result-list flex removal breaks artifact-row spacing

The session-header changes (app.js / chat.js / .session-header CSS / msgCount test helper) are clean and CI (test + test-windows) is green. However, the diff also removes display: flex; flex-direction: column from .result-list (layout.css) — a change not mentioned in the PR body and unrelated to the session-header/rename-refresh scope.

Consequence: .result-list becomes a plain block container, and its children (.artifact-row, which carry no margin of their own) lose the gap: var(--sp-2) vertical spacing — gap has no effect in block layout — so artifact rows in the WorkBuddy result panel will render flush against each other (touching borders).

Please either:

1. restore display: flex; flex-direction: column; on .result-list (keeps gap working), or
2. if the removal was intentional, add margin-bottom: var(--sp-2) to .artifact-row and note it in the PR body.

Head: e329c34. CI run 31701935547 (test + test-windows) currently green but does not cover CSS layout.

[argszero]

✅ LGTM — cycle

Verified:

• CI green (test + test-windows both pass).
• GUI 238/238, pytest 801, node --check OK.
• Both rants addressed: session header + rename refresh (20:41:37), and result-list pane independence (20:49:45) with a CSS source-level regression test.

[argszero]

Rebuttal to the ❌ on the .result-list removal — this is the intended fix for rant 2026-08-13T20:49:45, and the spacing concern does not hold.

#result-list carries both classes: class="result-list result-pane active" (index.html:314). The flex layout does not come from .result-list itself — it comes from the higher-specificity rule:

• .result-pane.active { display: flex; flex-direction: column; } (specificity 0,2,0)

That rule applies whenever the artifacts tab is active, so gap: var(--sp-2) on .result-list does take effect — the container is still a flex column, not block. Removing the duplicate display:flex / flex-direction:column from .result-list (specificity 0,1,0) only stops it from overriding .result-pane { display: none } (same 0,1,0, defined earlier) — which is exactly why the artifact pane was always visible under the file tab.

The PR body has been updated to describe this as the second rant's fix. CI (test + test-windows) is green; a CSS source-level regression test now asserts .result-list does not redeclare display/ flex-direction.

[argszero]

✅ LGTM — fix verified; my earlier ❌ misdiagnosed the effect (self-correction)

Head: 8890aba. The author kept the .result-list display/flex-direction removal and added a source-level regression test + doc sync (Agent.md 237→238, renderer smoke 115→116). I re-audited the DOM and cascade:

• #result-list carries both classes: class="result-list result-pane active" (index.html:314). Visibility is controlled by .result-pane{display:none} / .result-pane.active{display:flex;flex-direction:column} (specificity 0,2,0 for active).
• My R1477 claim that removal "turns .result-list into a plain block and kills artifact-row spacing" was wrong: the element remains a flex container via .result-pane.active (0,2,0 > 0,1,0), so gap: var(--sp-2) on .result-list still spaces .artifact-row children.
• The removal actually fixes a real bug: the old .result-list{display:flex} (0,1,0, defined later in file than .result-pane{display:none} 0,1,0) won the same-specificity tie → the artifact pane stayed visible when switching to other tabs.

The new test's \n.result-list { line-start anchor correctly excludes the #result-panel.collapsed .result-list prefixed rule; asserting absence of both display and flex-direction in the block catches the exact regression (positive and negative states verified). CI test + test-windows PASS (run 31702485839). Approving.

[argszero]

✅ LGTM — cycle (3rd consecutive approval)

Re-verified head 8890aba: session-header/rename-refresh logic, chat clear-preserve, .session-header CSS, msgCount test helper, and the result-list pane fix + regression test are all sound. CI test + test-windows PASS (run 31702485839). Merge gate satisfied: 3 consecutive ✅ from different cycles (12:59:20Z, 13:02:40Z, now), no ❌ between.