AlterMundi · nicoechaniz · Aug 7, 2026 · Aug 6, 2026 · Aug 6, 2026 · Aug 6, 2026
diff --git a/docs/ops/rehearsals/2026-08-06-post-aug2-human-qa.md b/docs/ops/rehearsals/2026-08-06-post-aug2-human-qa.md
@@ -0,0 +1,213 @@
+# HUMAN-QA-02 — Post-August-2 human acceptance sheet — 2026-08-06
+
+Issue: https://github.com/AlterMundi/harmonic-beacon-webapp/issues/204
+
+**Status: PREPARED — NOT EXECUTED**
+**Decision: PENDING DEPLOYED SHA + HUMAN EXECUTION**
+
+This sheet is **additive**. The canonical protocol is
+[`../WEEKEND_REHEARSAL.md`](../WEEKEND_REHEARSAL.md) — entry gate, role
+discipline, synthetic-identity rules, device matrix minimums and the rehearsal
+sequence all apply unchanged. This document only adds what landed after the
+August 2 free event: public questions-and-emotions contributions
+(NAMED/ANONYMOUS), the cockpit contributions drawer, the second-session A→B
+leakage check, and the bilingual SessionGuidance surface. Nothing here
+rewrites the protocol.
+
+No result in this sheet may be filled by delegation, by CI, or by reading the
+code. Every row requires a human executing against the deployed candidate and
+recording what they observed.
+
+## 1. Candidate identity
+
+| Field | Value |
+|---|---|
+| `origin/main` SHA at evidence refresh (2026-08-07) | `aec2632` |
+| Deployed SHA (`/api/health` `gitSha`, build 2026-08-07T00:17:59Z) | `64634e94f21a0a0d60d0f5745c159fe224f2895b` (= `origin/release`) |
+| Database schema (deployed) | `20260806233000_add_logos_private_session` |
+| Branch comparison | `git rev-list --left-right --count origin/release...origin/main` → `2 21`: **not** "main is 21 ahead" — the two branches carry their own commits because LOGOS was integrated separately on each: release via PR #209 (`71d996a` + merge `64634e9`), main via PR #208 (`d0364db` + merge `aec2632`) |
+| Rollback SHA | PENDING — record from the deploy workflow before execution |
+| URL / environment | `https://live.harmonicbeacon.com` (production event stack) |
+| Test session UUID | PENDING — derive from the canonical catalog/cockpit at execution time; never copy from historical docs |
+| Execution date/time (UTC) | PENDING |
+
+Verified facts at evidence refresh (2026-08-07, after the LOGOS release) —
+`/api/health` and `/api/health/ready` plus
+`git merge-base --is-ancestor <sha> 64634e94f…`, where **exit code 0 means
+"contained" and exit code 1 means "not contained"**:
+
+- `89f77991` (PR #161, contributions contract/backend) → **exit 0**:
+  production contains the contributions backend.
+- `388fe8f` (PR #187, attendee/staff UI) → **exit 1**: not deployed.
+- `671cde8` (final contribution hardening) → **exit 1**: not deployed.
+- `d547e96` (PR #192, E2E fixes) → **exit 1**: not deployed.
+- `341c15c` (PR #206, SessionGuidance) → **exit 1**: not deployed.
+- `git rev-list --left-right --count origin/release...origin/main` → `2 21`:
+  divergent branches with own LOGOS integration commits on each side (see
+  table above).
+
+**Fail-closed rule:** the new release `64634e9` adds the private LOGOS
+session, but production still contains only the contributions
+contract/backend of PR #161 — not the attendee/staff UI of PR #187, its
+final hardening, the PR #192 fixes, or SessionGuidance of PR #206. If
+`/api/health` at execution time does not show a candidate containing PRs
+#187, #192 **and #206**, mark every row **BLOCKED** and attribute no result
+to the new design. Execution is equally blocked until real humans fill every
+role in §2 — a prepared sheet is not an executed rehearsal.
+
+## 2. Roles (synthetic codes only)
+
+Never record names, emails, tickets or tokens in this sheet. Use only these
+codes:
+
+| Role code | Responsibility | Account/capability |
+|---|---|---|
+| GO-NOGO-OWNER | Go/no-go signature and abort authority | — |
+| FACILITATOR-01 | Facilitates the session | Assigned `FACILITATOR_OP` |
+| OP-01 | Independent operator | `OPERATOR` or `ADMIN` |
+| ATT-CHROME | Attendee, desktop Chrome | Synthetic `COMP` ticket |
+| ATT-FIREFOX | Attendee, desktop Firefox, different network | Synthetic `COMP` ticket |
+| ATT-IPHONE | Attendee, physical iPhone Safari | Synthetic `COMP` ticket |
+| ATT-ANDROID | Attendee, physical Android Chrome | Synthetic `COMP` ticket |
+| RECORDER-01 | Reads the script, records evidence, never coaches | No production credentials |
+
+## 3. Device and network matrix
+
+| Tester code | Role | Device | OS/version | Browser/version | Network | Headphones/speaker | Result | Evidence |
+|---|---|---|---|---|---|---|---|---|
+| ATT-CHROME | Attendee | PENDING | PENDING | Chrome PENDING | AR network | PENDING | NOT RUN | PENDING |
+| ATT-FIREFOX | Attendee | PENDING | PENDING | Firefox PENDING | Different network | PENDING | NOT RUN | PENDING |
+| ATT-IPHONE | Attendee | Physical iPhone PENDING | iOS PENDING | Safari PENDING | PENDING | PENDING | NOT RUN | PENDING |
+| ATT-ANDROID | Attendee | Physical Android PENDING | Android PENDING | Chrome PENDING | PENDING | PENDING | NOT RUN | PENDING |
+| ATT-TURN | Any attendee role | PENDING | PENDING | PENDING | Restrictive path (ICE/TCP or TURN) | PENDING | NOT RUN | PENDING |
+| ATT-BGFG | Mobile attendee | PENDING | PENDING | PENDING | Background/foreground cycle | PENDING | NOT RUN | PENDING |
+| ATT-BT | Mobile attendee where possible | PENDING | PENDING | PENDING | Bluetooth or output switching | Bluetooth PENDING | NOT RUN | PENDING |
+
+## 4. Test journey
+
+The canonical sequence from `WEEKEND_REHEARSAL.md` §1–§5 applies in full. The
+rows below make the post-August-2 additions explicit and give every row a
+traceable ID. All rows start **NOT RUN** with no observed result.
+
+| ID | Precondition | Action | Expected result | Observed | Result | Severity | Evidence | Issue |
+|---|---|---|---|---|---|---|---|---|
+| QA-01 | Doors open; tester has no briefing | ATT-CHROME explains their role, the session, and the next action without coaching | Correct, unaided comprehension | | NOT RUN | P2 | | |
+| QA-02 | In room | Tester states who they are and which session they are in | Identity/role/session named correctly | | NOT RUN | P0 | | |
+| QA-03 | In room | Deny microphone, allow camera (and vice versa) | Partial permissions handled with clear state, no dead end | | NOT RUN | P1 | | |
+| QA-04 | Audio active | Distinguish Stage, audience, tapestry and Beacon | All four identified correctly | | NOT RUN | P2 | | |
+| QA-05 | Audio active | Listen 60 s at balance 0.25 | Beacon/voice presence as designed; record glitches/routing | | NOT RUN | P1 | | |
+| QA-06 | Audio active | Listen 60 s at balance 0.50 | As QA-05 | | NOT RUN | P1 | | |
+| QA-07 | Audio active | Listen 60 s at balance 0.75 | As QA-05 | | NOT RUN | P1 | | |
+| QA-08 | Audio active, balance fully to Beacon | Move balance fully toward Beacon | Session voice at zero; only the Beacon is heard | | NOT RUN | P1 | | |
+| QA-09 | Session voice present | Human judges voice intelligibility | Voice intelligible (continuity alone is not intelligibility) | | NOT RUN | P1 | | |
+| QA-10 | Camera on, mic on | Turn camera off | Microphone and active audio continue; each control works independently | | NOT RUN | P0 | | |
+| QA-11 | Mobile, on Stage | Switch front/rear camera and back | No room exit, no mic/Beacon interruption | | NOT RUN | P1 | | |
+| QA-12 | In audience | Raise hand | Facilitator and operator both see it with name/thumbnail | | NOT RUN | P1 | | |
+| QA-13 | Hand raised | Staff view shows attendee name and thumbnail | Real name/thumbnail, not a generic role | | NOT RUN | P1 | | |
+| QA-14 | Hand raised | FACILITATOR-01 invites; attendee accepts | Publishes only after explicit accept; audience until then | | NOT RUN | P0 | | |
+| QA-15 | Hand raised | Invite; attendee declines | Never appears on Stage | | NOT RUN | P0 | | |
+| QA-16 | On Stage | Demote attendee | Stage, tapestry, hand queue and grants converge | | NOT RUN | P1 | | |
+| QA-17 | Was on Stage; network cut 15 s | Restore network and reconnect | No stale `ON_STAGE`; no automatic reappearance | | NOT RUN | P1 | | |
+| QA-18 | Tapestry visible | Observe tapestry and audience presence | Composition names participants; presence truthful | | NOT RUN | P2 | | |
+| QA-19 | In room, clean feed | ATT-CHROME posts a NAMED contribution | Appears immediately with the display name | | NOT RUN | P1 | | |
+| QA-20 | In room | ATT-FIREFOX posts an ANONYMOUS contribution | Audience sees "Anónimo"/"Anonymous", never the author | | NOT RUN | P0 | | |
+| QA-21 | QA-20 done | Every attendee device inspects the anonymous message | No author identity anywhere in the audience surface | | NOT RUN | P0 | | |
+| QA-22 | QA-20 done | FACILITATOR-01/OP-01 opens the staff feed | Real author visible to staff with anonymity badge | | NOT RUN | P0 | | |
+| QA-23 | Composer open | Press Enter inside the composer | New line only; nothing publishes implicitly | | NOT RUN | P1 | | |
+| QA-24 | Draft written; network cut | Attempt send while offline | Draft preserved; recovery offered; no duplicate on retry | | NOT RUN | P1 | | |
+| QA-25 | Backlog open | New messages arrive while reading backlog | "New messages ↓" indicator; no forced scroll | | NOT RUN | P2 | | |
+| QA-26 | Audio active with media probe (where available) | Open/close the contributions panel and drawer | No remount; no new sockets, tracks, media elements or AudioContext | | NOT RUN | P0 | | |
+| QA-27 | Audio active with media probe | Open/close SessionGuidance | No remount; media pipeline untouched | | NOT RUN | P0 | | |
+| QA-28 | Guidance closed | ATT-CHROME reads SessionGuidance in ES | Explains volume, balance, full-Beacon listening and camera/mic independence unaided | | NOT RUN | P2 | | |
+| QA-29 | Guidance closed | A second attendee reads it in EN | Same comprehension in English | | NOT RUN | P2 | | |
+| QA-30 | Guidance visible pre-room | Tester waiting behind closed doors finds and reads the guidance | Available and understandable before entry | | NOT RUN | P2 | | |
+| QA-31 | Session A done | Move the same principal to session B | No leakage of messages, cursors, drafts, hand, grant, language, audio or stage state | | NOT RUN | P0 | | |
+| QA-32 | Session live | FACILITATOR-01 ends the session from the cockpit | All attendees exit; no new tokens issued | | NOT RUN | P0 | | |
+| QA-33 | QA-32 done | Staff and attendees observe final state | Converged final state; closure communicated as completion, not transport error | | NOT RUN | P1 | | |
+| QA-34 | QA-32 done | A former attendee re-opens the session URL | Entry denied; no token issued; truthful ended state | | NOT RUN | P1 | | |
+
+## 5. Human ES/EN language review
+
+Run Parts A and B of
+[`../../verification/HUMAN_ROLE_LANGUAGE_REVIEW.md`](../../verification/HUMAN_ROLE_LANGUAGE_REVIEW.md)
+against the same deployed SHA. The review must answer, unaided, in both
+locales:
+
+- Who are you in this room?
+- What can you do here?
+- What can you not do?
+- Who controls your camera and microphone?
+- What does raising your hand imply?
+- How does anonymity work when you share?
+- What does the overall volume do?
+- What does the Beacon/Session balance do?
+- What does "look for a question, not an answer" mean to you?
+
+Do not fill answers on behalf of testers. Record per-locale outcomes in the
+review's own evidence sheet and link it here: **PENDING**.
+
+## 6. Qualitative interview
+
+Record without steering answers:
+
+| Field | Recording |
+|---|---|
+| Moment of greatest confusion | PENDING |
+| Action that felt scary | PENDING |
+| Expectation when turning the camera off | PENDING |
+| Perceived meaning of "Beacon" | PENDING |
+| Feature discovered too late | PENDING |
+| Confidence in anonymity | PENDING |
+| Visual hierarchy reading | PENDING |
+| Room versus dashboard perception | PENDING |
+| Copy problem found | PENDING |
+| Touch/layout problem found | PENDING |
+
+## 7. Severity and abort
+
+- **P0:** wrong identity or session; consent/privacy breach; anonymous author
+  exposed; duplicated audio/media; seventh publisher; inability to end the
+  session.
+- **P1:** main path unusable; unintelligible voice; wrong reconnect/Stage
+  state; audio loss; inconsistent contribution behavior.
+- **P2:** ambiguous copy; problematic layout/touch; hard-to-find feature.
+- **P3:** visual refinement with no operational impact.
+
+**Abort conditions:** any P0; health degradation on the target; unexpected
+production impact; inability to close the session. On abort, stop, record the
+state, and notify the go/no-go owner. Open a minimal issue per real defect;
+never silently edit evidence.
+
+## 8. Scope separation
+
+- This battery does **not** certify 150 participants and does not close #99.
+- It does **not** test payments and does not close #25/#26/#81 (existing
+  commercial evidence from Mariano/Sai is linked, not re-executed).
+- It does **not** test moderation/withdrawal: #134 remains pending; register
+  its absence as a pending feature, not a failure.
+- It does **not** execute or certify the facilitation-driven shared mix
+  override: **#98 (RTC-02) exists and remains OPEN and unimplemented**; that
+  work stays under `audio-touching` governance and requires explicit acoustic
+  approval. No interface copy claims the override exists.
+- It does **not** require a mic checker: #126 remains pending.
+- It does **not** require staff audio return: #131 remains pending.
+- EarlyBirds (PR #203 line) is a separate staging lane: not part of this
+  battery, not a blocker, not touched.
+- This battery authorizes no sales and no deploy.
+
+## 9. Closure
+
+| Field | Value |
+|---|---|
+| PASS / FAIL / BLOCKED / NOT RUN summary | PENDING |
+| Issues created | PENDING |
+| Open P0/P1 | PENDING |
+| Decision | PENDING — GO / CONDITIONAL GO / NO-GO |
+| Go/no-go owner signature | PENDING |
+| Linked from #24, #64, #66, #68, #69, #94 | PENDING |
+
+Silence or incomplete rows are **not** approval. A row without human evidence
+stays NOT RUN; a candidate mismatch makes every row BLOCKED. The go/no-go
+decision requires the deployed SHA to contain PRs #187, #192 and #206, every
+P0/P1 closed or explicitly waived by the go/no-go owner, and this sheet linked
+from #24, #64, #66, #68, #69 and #94.