Session Closeout · ClientsFlow Pipeline · 2026-06-25

Everything you need to close this chat confidently

Done · what to read · every pending task · the prompt for a fresh Claude instance

NOT PUSHED main 2d79eba · pipeline v111 · 900 green · 167 ahead of origin · Studio on -rebuild only (not promoted) · 6 real deals safe

✅ What got done

📖 What you should read

🟥 Pending — P0 (the main unfinished task)

The EBO×Visual-QA toggle HTML the previous chat was interrupted on. Two-phase; the first deliverable is just two layout mockups for you to choose between (no test run).

🟧 Pending — P1 (needs your decision / trigger)

🟨 Pending — P2 (follow-ups / residuals)

📂 The six source EBOs (for the re-merge)

SourceRowsPathNote
studio-pipeline-integration13 (S1–S12+INV)plans/studio-pipeline-integration/{EBO.md, spec.json}FRESHEST for Studio
extended-ebo-2026-06-259 (B1–B6)plans/extended-ebo-2026-06-25/index.htmltouchpoint history; source is the HTML
bnf-2026-06-24-r312 (I1–I14)plans/bnf-2026-06-24-r3/{EBO.md, spec.json}round-3 pipeline fixes
bnf-2026-06-23-duo214 (BUG-01..11)plans/bnf-2026-06-23-duo2/{EBO.md, spec.json}duo-2 live-test bugs
board-card-redesign17plans/board-card-redesign/spec.jsonboard/stage/card behaviors
live-test-bug-fixes11plans/live-test-bug-fixes/spec.jsonpost-call → sign → pay
MASTER_EBO already merged these into 55 deduped rows (master_spec.json). Phase 2 must EXPAND back so every original scenario is present and conflicts are highlighted — the dedup hid them.

🤖 Prompt for a fresh Claude Code instance

Paste this into a new session to continue the P0 task. (Also in the handoff doc.)

Resume the ClientsFlow Pipeline work — I closed the previous chat. Read these FIRST, in order: 1. plans/whole-journey-qa-2026-06-25/CLOSEOUT_HANDOFF.md (the full state + the task spec — START HERE) 2. plans/SESSION_HANDOFF_2026-06-25.md (top banner "UPDATE 21:30") 3. plans/whole-journey-qa-2026-06-25/final_qa.json + MASTER_EBO.md + master_spec.json 4. the visual-qa-ultra report structure: ~/.claude/skills/visual-qa-ultra/scripts/report.py (the per-frame model: honesty tag -> "why/what I expect" -> screenshot -> (1) Gemini 5-sentence -> (2) Claude verify [Expected/area, What I see, Verdict, Next action] -> optional DECISION -> error-arc) State: local main 2d79eba, pipeline deployed v111, 900 pytest green, 167 commits ahead of origin (NOTHING pushed). Studio integration PROVEN LIVE (WJ-41..45); Studio on -rebuild only, NOT promoted to -prod. YOUR TASK (P0) — build the EBO x Visual-QA toggle HTML, in TWO PHASES: PHASE 1 (do this FIRST, run NO browser test): produce TWO different LAYOUT MOCKUPS of one HTML template, using only the first 3 scenarios (WJ-01/02/03) or dummy data, then STOP and let me pick one. The template is a whole-journey EBO where: - Columns = the MASTER_EBO columns EXCEPT the UX-critique column, i.e. ID | Behavior(=expected) | Strongest seam | (clock) Touchpoint history | QA bucket | Sources - PLUS test columns folded in from visual-qa-ultra: honesty tag + live model reasoning (Gemini 5-sentence + Claude verify) + verdict. - EACH scenario is a TOGGLE that expands to its TEST PANEL: the screenshots for that test + the live reasoning (the visual-qa-ultra per-frame block). Layout A = table with expandable rows; Layout B = card accordion (visual-qa-ultra-style vertical flow). Use .tmp/qa-shots/zz-card-booked.png + zz-card-archived.png as the dummy screenshots. House style: ~/.claude/house-style/report.css. Open both mockups locally + deploy to Cloudflare; give me the two URLs side by side so I can choose. PHASE 2 (only after I pick): RE-MERGE all six source EBOs into the chosen layout so it contains EVERY scenario that appeared in ANY prior EBO (this is a FULL merge, NOT the dedup MASTER_EBO did). Use the FRESHEST version of each scenario; where two EBOs disagree on a scenario's expected behavior, HIGHLIGHT it as a merge conflict. Sources (freshest noted in the handoff): plans/studio-pipeline-integration/{EBO.md,spec.json} (13) ; plans/extended-ebo-2026-06-25/index.html (9) ; plans/bnf-2026-06-24-r3/{EBO.md,spec.json} (12) ; plans/bnf-2026-06-23-duo2/{EBO.md,spec.json} (14) ; plans/board-card-redesign/spec.json (17) ; plans/live-test-bug-fixes/spec.json (11). Start from master_spec.json (55 deduped rows, sources cited) as the spine, but EXPAND back to every original scenario + flag conflicts. The test panels stay as scaffolding/placeholders — do NOT run visual-qa-ultra in Phase 2 (the live test is a separate later step). RULES: orchestrate-only (delegate non-trivial code to a bounded TDD subagent off a verified worktree from current main); NO push; never promote Studio to -prod; ZZ canonical sentinels only; never touch the 6 real deals; Notion archive-never-delete; human gate intact; browser = Chrome DevTools MCP or Playwright (never Puppeteer, never headless); reviewable docs = house-style HTML opened locally + deployed to Cloudflare with the URL in chat. Reply in the user's language (HU->HU, EN->EN).