대시보드 폴드아웃/드릴다운 정리 + 페르소나 역린·misconduct 반응 + 게이트웨이 격리·RAG 비차단 수정

SSOT 대시보드: - 한신대 기술분석 PDF(19쪽) 정합성 분석 + 이번 세션 발견 섹션 추가 - 섹션 폴드아웃(접기)·상단 목차(드릴다운)·모두 펼치기/접기 — 내용 보존, 레이아웃만 정리 페르소나 반응 강화('저항·반응 조절' 핵심 차별): - PersonaCard.triggers(역린) 필드 + CCD 핵심상처 파생 역린 블록 - L0에 무례·모욕·조롱 시 현실적 동맹 균열 반응 지침 버그·성능 수정(라이브/E2E로 포착): - 게이트웨이 페르소나 격리: --append-system-prompt를 --system-prompt(교체)로 + --exclude-dynamic-system-prompt-sections (내담자 캐릭터 붕괴·개발맥락 누출 차단) - RAG: 임베더 동기 로드(약 7-13초)를 _warm_rag_caches 백그라운드 warm으로(세션 생성 블로킹 회귀 수정) - voice TTS RMS 데드힌트 제거, init_state OpennessParams 파라미터객체화 - 한국어 PII(날짜·금액·주소) 마스킹 보강 - 레이아웃 시각 게이트: 폼 컨트롤 값 스크롤 오탐 제외(7/7) 검증: 백엔드 84/84, E2E 42(데스크톱 27·모바일 11·아바타 4), 시각 게이트 7/7
2026-06-27 02:30:46 +09:00 · 2026-06-27 02:30:46 +09:00 · 085460b5e0
commit 085460b5e0
parent cb2aebd76c
327 changed files with 31226 additions and 1829 deletions
--- a/docs/phase3/README.md
+++ b/docs/phase3/README.md
@ -0,0 +1,58 @@
+# Phase 3 Operational Gates
+
+This folder turns the Phase 3 dashboard items into executable operator gates for the
+20-person education pilot. It is not a legal opinion; the privacy and consent gates
+must be reviewed by the project data steward and counsel before real participant data
+is collected.
+
+## Artifacts
+
+- [pilot-runbook.md](pilot-runbook.md) - roster, session completion, operator cadence,
+  and evidence files for the 20-person education pilot.
+- [kpi-measurement-plan.md](kpi-measurement-plan.md) - KPI definitions, thresholds,
+  data sources, and report shape.
+- [dataset-export-manifest.md](dataset-export-manifest.md) - anonymized JSONL export
+  and manifest expectations for recursive-learning dataset output.
+- [privacy-consent-audit.md](privacy-consent-audit.md) - consent, withdrawal, privacy,
+  and audit checklist.
+
+## Evidence Root
+
+Operators should keep real pilot evidence outside the source tree by default, for
+example:
+
+```powershell
+$env:PHASE3_EVIDENCE_ROOT="D:\vignette-evidence\phase3-pilot-YYYYMMDD"
+python scripts\check-phase3-artifacts.py --check --evidence-root $env:PHASE3_EVIDENCE_ROOT
+```
+
+The expected evidence root layout is:
+
+```text
+00-intake/
+  pilot_roster.csv
+  consent_receipts.csv
+01-sessions/
+  session_completion.csv
+02-measures/
+  prepost_measures.csv
+  sus_responses.csv
+  kpi_report.json
+03-export/
+  export_manifest.json
+04-privacy/
+  privacy_audit.md
+  withdrawal_log.csv
+```
+
+No evidence file should contain names, email addresses, phone numbers, raw student IDs,
+raw source case records, raw audio, or third-party API secrets. Participant identity
+mapping, if required, must live in the approved restricted location owned by the data
+steward, not in the export or this repository.
+
+## Phase 3 Dashboard Coverage
+
+- 20-person education pilot: `pilot-runbook.md`
+- KPI measurement: `kpi-measurement-plan.md`
+- Recursive-learning dataset export: `dataset-export-manifest.md`
+- Privacy, consent, withdrawal audit: `privacy-consent-audit.md`
--- a/docs/phase3/dataset-export-manifest.md
+++ b/docs/phase3/dataset-export-manifest.md
@ -0,0 +1,175 @@
+# Phase 3 Anonymized Dataset Export Manifest
+
+## Purpose
+
+Define the manifest and JSONL expectations for the first recursive-learning dataset
+artifact. The export is allowed only after consent scope, withdrawal state, PII masking,
+and reviewer agreement are documented.
+
+Primary evidence file: `03-export/export_manifest.json`
+
+## Export Gate
+
+The export may be marked `approved_for_recursive_learning_seed` only when:
+
+- Source participants are active or otherwise allowed by consent scope at export time.
+- Withdrawn participants are excluded according to the approved withdrawal policy.
+- Direct identifiers and raw identity mapping are absent from exported files.
+- Free text is masked or transformed according to the manifest.
+- PII scan is pass, or every finding has a reviewer disposition.
+- IAA gates pass for gold exports: kappa >=0.60 and ICC >=0.75.
+- Legal/data-steward review is complete.
+
+If any gate is missing, set `export_status` to `technical_dry_run` or `blocked`, not
+`approved_for_recursive_learning_seed`.
+
+## Manifest Shape
+
+```json
+{
+  "export_id": "phase3-rl-seed-YYYYMMDD-001",
+  "dataset_name": "vignette_phase3_recursive_learning_seed",
+  "export_status": "technical_dry_run",
+  "created_at": "2026-06-26T00:00:00Z",
+  "purpose": "recursive-learning seed dataset for education simulator improvement",
+  "source_window": {
+    "started_at": "2026-00-00T00:00:00Z",
+    "ended_at": "2026-00-00T00:00:00Z"
+  },
+  "source_tables": [
+    "app.sessions",
+    "app.turns",
+    "app.feedback_scores",
+    "ds.annotation",
+    "audit.persona_drift_log"
+  ],
+  "selection_criteria": {
+    "cohort_id": "phase3",
+    "min_completed_sessions": 2,
+    "include_withdrawn": false,
+    "excluded_safety_scope": ["self_harm_scenario_primary"]
+  },
+  "consent_scope": {
+    "consent_version": "",
+    "allowed_uses": ["education_quality_review", "recursive_learning_seed"],
+    "withdrawal_cutoff_applied_at": "",
+    "participants_included": 0,
+    "participants_excluded": 0
+  },
+  "anonymization": {
+    "participant_key": "pseudonymous export key; no identity map included",
+    "text_transform": "masked_text_only",
+    "direct_identifier_policy": "blocked",
+    "salt_or_identity_map_location": "not in export"
+  },
+  "pii_scan": {
+    "tool": "",
+    "version": "",
+    "ran_at": "",
+    "status": "pending",
+    "findings": []
+  },
+  "agreement": {
+    "kappa": null,
+    "icc": null,
+    "gold_status": "not_gold"
+  },
+  "files": [
+    {
+      "path": "03-export/anonymized_dataset.jsonl",
+      "rows": 0,
+      "sha256": "",
+      "schema": "phase3_dataset_item_v1"
+    }
+  ],
+  "approvals": {
+    "data_steward": "",
+    "legal_or_privacy_reviewer": "",
+    "technical_operator": "",
+    "approved_at": ""
+  },
+  "known_limitations": []
+}
+```
+
+## JSONL Record Shape
+
+Each JSONL line must be one dataset item:
+
+```json
+{
+  "schema": "phase3_dataset_item_v1",
+  "item_id": "export-local-id",
+  "participant_key": "PX-0001",
+  "session_key": "SX-0001",
+  "turn_key": "TX-0001",
+  "persona_id": "P1",
+  "stage": "rapport",
+  "speaker": "counselor",
+  "text_masked": "masked or transformed utterance text",
+  "techniques": [],
+  "client_states": [],
+  "feedback_scores": [],
+  "supervisor_comments": [],
+  "source_refs": {
+    "session_started_at": "2026-00-00T00:00:00Z",
+    "export_manifest_id": "phase3-rl-seed-YYYYMMDD-001"
+  },
+  "privacy": {
+    "direct_identifiers_removed": true,
+    "pii_scan_status": "pass",
+    "consent_scope": "recursive_learning_seed"
+  }
+}
+```
+
+Blocked fields:
+
+- name
+- email
+- phone
+- student_id
+- national_id
+- address
+- date_of_birth
+- raw_audio_path
+- raw_voice
+- raw_source_case
+- identity_map
+- API keys, tokens, cookies, or credentials
+
+## Operator Commands
+
+Generate the export only through an approved read-only reporting job. Until that job is
+implemented, use the manifest as a contract and mark exports `technical_dry_run`.
+
+Validate artifact presence and headers:
+
+```powershell
+python scripts\check-phase3-artifacts.py --check --evidence-root $root
+```
+
+Write a validation report:
+
+```powershell
+python scripts\check-phase3-artifacts.py --check --evidence-root $root --output "$root\phase3_artifact_check.json"
+```
+
+Compute a file hash for the manifest:
+
+```powershell
+Get-FileHash "$root\03-export\anonymized_dataset.jsonl" -Algorithm SHA256
+```
+
+## Export Review Checklist
+
+- [ ] Manifest `export_status` matches the actual approval state.
+- [ ] Files listed in `files[]` exist and have row counts and SHA256 hashes.
+- [ ] `participants_included` plus `participants_excluded` matches the consent roster
+  decision log.
+- [ ] `include_withdrawn` is false unless legal review explicitly approved otherwise.
+- [ ] Text fields use masked/transformed text only.
+- [ ] PII scan report is attached or referenced.
+- [ ] Agreement metrics match `02-measures/kpi_report.json`.
+- [ ] Export limitations state whether the file is gold data or a dry-run seed.
+
--- a/docs/phase3/kpi-measurement-plan.md
+++ b/docs/phase3/kpi-measurement-plan.md
@ -0,0 +1,165 @@
+# Phase 3 KPI Measurement Plan
+
+## Purpose
+
+Define the Phase 3 KPI report so the dashboard gates are measured the same way each
+time. These metrics are pilot evidence, not claims of validated clinical effectiveness.
+
+Primary evidence file: `02-measures/kpi_report.json`
+
+## Report Shape
+
+```json
+{
+  "pilot_id": "phase3-pilot-YYYYMMDD",
+  "generated_at": "2026-06-26T00:00:00Z",
+  "source_window": {
+    "started_at": "2026-00-00T00:00:00Z",
+    "ended_at": "2026-00-00T00:00:00Z"
+  },
+  "cohort_size": 20,
+  "metrics": {
+    "pilot_completion": {},
+    "session_completion": {},
+    "embedding_consistency": {},
+    "hallucination_rate": {},
+    "top1": {},
+    "sus": {},
+    "inter_rater_kappa": {},
+    "icc": {},
+    "self_efficacy_prepost": {}
+  },
+  "exclusions": [],
+  "open_schema_gaps": [],
+  "review": {
+    "operator": "",
+    "reviewed_at": "",
+    "decision": "pending"
+  }
+}
+```
+
+Each metric object must include:
+
+```json
+{
+  "value": 0,
+  "threshold": 0,
+  "pass": false,
+  "numerator": null,
+  "denominator": null,
+  "method": "",
+  "source_files": []
+}
+```
+
+## KPI Definitions
+
+| KPI | Gate | Definition | Evidence |
+| --- | --- | --- | --- |
+| Pilot completion | >=20 participants | Count active participants with consent, required sessions, pre/post, SUS, and supervisor review. | `pilot_roster.csv`, `session_completion.csv`, `prepost_measures.csv`, `sus_responses.csv` |
+| Session completion | >=20 participants | Count participants with at least two `completed` sessions and non-empty `supervisor_reviewed_at`. | `session_completion.csv` |
+| Embedding consistency | >=0.79 | Mean cosine similarity for persona invariant checks across masked session turns. Report model, embedding version, mean, p10, and count. | DB query export, `kpi_report.json` |
+| Hallucination rate | <=0.05, target <=0.01 | Unsupported or contradicted factual claims divided by reviewed eligible evaluator/client-response claims. | supervisor review export, `kpi_report.json` |
+| Top-1 retrieval | >=0.80 | Top-1 evidence match rate for golden feedback or RAG evaluation queries. | golden evaluation run, `kpi_report.json` |
+| SUS | >=80 | Standard 10-item SUS score transformed to a 0-100 score; report mean, median, and response count. | `sus_responses.csv` |
+| Cohen kappa | >=0.60 | Agreement between human reviewers or human-vs-AI labels for categorical tags used in dataset export. | annotation export, `kpi_report.json` |
+| ICC | >=0.75 | Intraclass correlation for numeric ratings such as feedback score dimensions. | annotation export, `kpi_report.json` |
+| Self-efficacy pre/post | Report and review | Paired pre/post change for the approved self-efficacy measure. Report effect direction, test used, p-value, and missingness. | `prepost_measures.csv` |
+
+## Metric Methods
+
+### Pilot Completion
+
+Computation:
+
+```text
+participant passes =
+  active consent receipt exists
+  AND >=2 completed sessions
+  AND every counted session has supervisor_reviewed_at
+  AND pre measure exists
+  AND post measure exists
+  AND SUS response set exists or missingness is explicitly reviewed
+```
+
+The dashboard gate passes when at least 20 participants pass. Withdrawn participants are
+reported in exclusions and must not be silently removed from denominator notes.
+
+### Embedding Consistency
+
+Operator command placeholder:
+
+```powershell
+# Use the deployed reporting job when available. The output is copied into kpi_report.json.
+python scripts\check-phase3-artifacts.py --check --evidence-root $root
+```
+
+Minimum report fields:
+
+```json
+{
+  "value": 0.79,
+  "threshold": 0.79,
+  "pass": true,
+  "method": "masked persona invariant cosine similarity",
+  "model": "embedding-model-name-and-version",
+  "p10": 0.0,
+  "count": 0,
+  "source_files": ["02-measures/persona_embedding_eval.csv"]
+}
+```
+
+### Hallucination Rate
+
+Count only reviewed claims that should be grounded in the persona card, session state,
+approved knowledge base, or rubric. Unsupported clinical, safety, or personal-history
+claims are high severity and must be listed in `review.exceptions`.
+
+Formula:
+
+```text
+hallucination_rate = unsupported_or_contradicted_claims / reviewed_eligible_claims
+```
+
+### SUS
+
+Report the standard SUS transformed 0-100 score. Do not copy item text into the
+repository. Store only item number, numeric response, participant id, and timestamp.
+
+Minimum report fields:
+
+```json
+{
+  "value": 80.0,
+  "threshold": 80.0,
+  "pass": true,
+  "mean": 80.0,
+  "median": 80.0,
+  "respondents": 20,
+  "source_files": ["02-measures/sus_responses.csv"]
+}
+```
+
+### Inter-Rater Agreement
+
+Use the same item universe for kappa and ICC that is later referenced by the dataset
+export manifest. If agreement fails, the dataset can still be exported as a technical
+artifact only when the manifest marks it `not_gold`.
+
+## Operator Checklist
+
+- [ ] Evidence root exists outside the source tree.
+- [ ] Roster and consent receipt counts match.
+- [ ] Session completion rows are exported after supervisor review.
+- [ ] Pre/post and SUS files are present.
+- [ ] `kpi_report.json` contains all required metric keys.
+- [ ] Every failed metric has an owner, disposition, and decision date.
+- [ ] Export manifest references the same cohort and exclusions as the KPI report.
+
+Validation command:
+
+```powershell
+python scripts\check-phase3-artifacts.py --check --evidence-root $root
+```
+
--- a/docs/phase3/pilot-runbook.md
+++ b/docs/phase3/pilot-runbook.md
@ -0,0 +1,187 @@
+# Phase 3 Education Pilot Runbook
+
+## Purpose
+
+Run a 20-person education pilot with auditable evidence that each included participant
+was consented, completed the required learning sessions, produced pre/post measures,
+and was included or excluded from KPI and export outputs correctly.
+
+This runbook assumes Phase 2 role separation, session persistence, supervisor review,
+and safety event logging are already deployed. It does not authorize real participant
+collection until the privacy and consent checklist is signed.
+
+## Entry Gates
+
+- `G-P3-01` Product scope is locked as education/training, not diagnosis or treatment.
+- `G-P3-02` Current consent form, privacy notice, retention rule, withdrawal path, and
+  third-party/inference transfer disclosure are approved by the project owner and legal
+  reviewer.
+- `G-P3-03` The pilot uses pseudonymous `participant_id` values only. Names, emails,
+  phone numbers, student numbers, and direct identifiers are not stored in evidence
+  files.
+- `G-P3-04` Safety policy is active: self-harm scenarios remain excluded unless a later
+  approved protocol explicitly adds them.
+- `G-P3-05` Instructor/admin access has RBAC and audit logging enabled before any pilot
+  session starts.
+- `G-P3-06` Evidence root has been created outside the source tree and the checker runs:
+
+```powershell
+$root="D:\vignette-evidence\phase3-pilot-YYYYMMDD"
+python scripts\check-phase3-artifacts.py --check --evidence-root $root
+```
+
+The first run is expected to fail on missing files. The run output becomes the setup
+gap list.
+
+## Participant Roster Gate
+
+Target: 20 active, consented trainees.
+
+Evidence file: `00-intake/pilot_roster.csv`
+
+Required header:
+
+```csv
+participant_id,cohort_id,consent_version,consent_signed_at,withdrawal_state,enrolled_at
+```
+
+Allowed `withdrawal_state` values:
+
+- `active`
+- `withdrawn_before_data_use`
+- `withdrawn_after_data_use`
+- `excluded_by_operator`
+
+Rules:
+
+- `participant_id` must be a pseudonymous value such as `P3-001`.
+- The roster must not include names, emails, phone numbers, student IDs, addresses, or
+  date of birth.
+- A separate identity map may exist only in the approved restricted location owned by
+  the data steward.
+- The pilot can start when 20 participants are `active` with a non-empty
+  `consent_signed_at`.
+
+Evidence file: `00-intake/consent_receipts.csv`
+
+Required header:
+
+```csv
+participant_id,consent_version,signed_at,signer_role,receipt_id
+```
+
+Operator command:
+
+```powershell
+python scripts\check-phase3-artifacts.py --check --evidence-root $root
+```
+
+## Session Completion Gate
+
+Minimum completion definition for dashboard Phase 3:
+
+- Consent receipt exists before the first session.
+- Participant has at least two completed simulation sessions.
+- Each completed session has an end timestamp, non-zero learner turns, and supervisor
+  review evidence.
+- Participant has both pre and post measures.
+- Participant has SUS responses or an explicit missing-response reason in the KPI
+  report.
+
+Evidence file: `01-sessions/session_completion.csv`
+
+Required header:
+
+```csv
+participant_id,session_id,persona_id,started_at,ended_at,completion_state,turns_count,supervisor_reviewed_at
+```
+
+Allowed `completion_state` values:
+
+- `completed`
+- `abandoned`
+- `operator_cancelled`
+- `excluded_from_analysis`
+
+Operational cadence:
+
+1. Before each session, confirm the participant remains `active`.
+2. After each session, export one row to `session_completion.csv`.
+3. Supervisor reviews the persisted session and records `supervisor_reviewed_at`.
+4. Any safety event or access override is linked in the KPI report, not copied into the
+   roster.
+
+Session evidence query placeholder:
+
+```sql
+-- Run against the approved read-only reporting connection.
+-- Replace table/column names with the deployed schema names if they differ.
+SELECT
+  s.learner_id AS participant_id,
+  s.id AS session_id,
+  s.persona_id,
+  s.started_at,
+  s.ended_at,
+  CASE WHEN s.ended_at IS NOT NULL THEN 'completed' ELSE 'abandoned' END AS completion_state,
+  COUNT(t.id) FILTER (WHERE t.speaker = 'counselor') AS turns_count,
+  MAX(r.reviewed_at) AS supervisor_reviewed_at
+FROM app.sessions s
+LEFT JOIN app.turns t ON t.session_id = s.id
+LEFT JOIN app.session_reviews r ON r.session_id = s.id
+GROUP BY s.learner_id, s.id, s.persona_id, s.started_at, s.ended_at;
+```
+
+If the deployed schema does not yet have `app.session_reviews`, the operator must
+attach a signed reviewer export and note the schema gap in `02-measures/kpi_report.json`.
+
+## Measure Collection Gate
+
+Evidence file: `02-measures/prepost_measures.csv`
+
+Required header:
+
+```csv
+participant_id,measure_name,timepoint,score,collected_at
+```
+
+Allowed `timepoint` values:
+
+- `pre`
+- `post`
+
+Evidence file: `02-measures/sus_responses.csv`
+
+Required header:
+
+```csv
+participant_id,item,response,collected_at
+```
+
+Rules:
+
+- `measure_name` must be stable across pre and post rows.
+- SUS rows use item numbers and numeric responses only. Do not store free-text comments
+  in `sus_responses.csv`.
+- Free-text participant feedback, if collected, must go through the privacy review path
+  before being added to any export.
+
+## Exit Gate
+
+Phase 3 pilot completion evidence is sufficient when:
+
+- `pilot_roster.csv` contains at least 20 active, consented participants.
+- At least 20 participants meet the minimum completion definition.
+- `kpi_report.json` is generated and reviewed.
+- `export_manifest.json` is generated, PII scan is pass or reviewed, and the export is
+  approved.
+- `privacy_audit.md` is signed with no unresolved high severity item.
+- `withdrawal_log.csv` exists, even if it only contains the header.
+
+Archive command:
+
+```powershell
+python scripts\check-phase3-artifacts.py --check --evidence-root $root --output "$root\phase3_artifact_check.json"
+```
+
+The checker writes only the report path given by `--output`; otherwise it is read-only.
+
--- a/docs/phase3/privacy-consent-audit.md
+++ b/docs/phase3/privacy-consent-audit.md
@ -0,0 +1,152 @@
+# Phase 3 Privacy, Consent, and Withdrawal Audit
+
+## Purpose
+
+Provide an operator checklist for consent, withdrawal, privacy review, and evidence
+collection before and during the 20-person education pilot. This is an operational
+artifact, not legal advice.
+
+Primary evidence files:
+
+- `04-privacy/privacy_audit.md`
+- `04-privacy/withdrawal_log.csv`
+
+## Pre-Pilot Privacy Gate
+
+Copy this checklist into `04-privacy/privacy_audit.md` and complete it with reviewer
+names, dates, and evidence links.
+
+```markdown
+# Phase 3 Privacy Audit
+
+Pilot ID:
+Evidence root:
+Review date:
+Operator:
+Data steward:
+Legal/privacy reviewer:
+
+## Scope
+- [ ] Intended use is education/training, not diagnosis, treatment, or emergency care.
+- [ ] Participant population and inclusion/exclusion criteria are approved.
+- [ ] Self-harm primary scenarios remain excluded unless separately approved.
+- [ ] Minor/source-case data use, if any, has documented lawful consent or approval.
+
+## Consent
+- [ ] Consent form version is recorded.
+- [ ] Consent covers session storage, review, KPI reporting, and any dataset export use.
+- [ ] Consent explains third-party model or inference transfer, if applicable.
+- [ ] Consent explains withdrawal process and practical limits after anonymized export.
+- [ ] Consent receipt is captured before first session.
+
+## Privacy Controls
+- [ ] Evidence files use pseudonymous participant IDs only.
+- [ ] Identity map location is restricted and not in the repository or export.
+- [ ] PII masking is active before external model calls.
+- [ ] Raw audio, raw source cases, and direct identifiers are excluded from Phase 3
+      artifacts unless separately approved.
+- [ ] Retention and deletion schedule is documented.
+- [ ] RBAC and audit logging are enabled for instructor/admin views.
+- [ ] Export PII scan process is documented.
+
+## Vendor and Transfer Review
+- [ ] Model providers and regions are listed.
+- [ ] Data sent to each provider is described at field level.
+- [ ] Secrets are not stored in evidence files.
+- [ ] Any overseas transfer, subprocessors, or institutional requirements have been
+      reviewed by counsel or the designated privacy reviewer.
+
+## Decision
+- [ ] Approved to start pilot.
+- [ ] Approved to generate KPI report.
+- [ ] Approved to generate anonymized export.
+- [ ] Blocked pending legal/privacy review.
+
+Open issues:
+```
+
+## Withdrawal Log
+
+Evidence file: `04-privacy/withdrawal_log.csv`
+
+Required header:
+
+```csv
+participant_id,requested_at,effective_at,scope,status,attestation_path
+```
+
+Allowed `scope` values:
+
+- `future_sessions_only`
+- `exclude_from_analysis`
+- `exclude_from_export`
+- `delete_where_policy_allows`
+
+Allowed `status` values:
+
+- `received`
+- `in_progress`
+- `completed`
+- `rejected_by_policy`
+- `needs_legal_review`
+
+## Withdrawal Runbook
+
+1. Record the request in `withdrawal_log.csv` using only `participant_id`.
+2. Stop new sessions for that participant.
+3. Mark the roster `withdrawal_state`.
+4. Identify sessions, turns, measures, review records, and export rows in scope.
+5. Exclude the participant from future exports unless legal review states otherwise.
+6. Remove or anonymize records where the approved retention/deletion policy allows.
+7. Create an attestation note at the path listed in `attestation_path`.
+8. Re-run the artifact checker and KPI report.
+9. If an export was already produced, add a manifest limitation and create a replacement
+   export if required by policy.
+
+Attestation template:
+
+```markdown
+# Withdrawal Attestation
+
+Participant ID:
+Request received:
+Effective date:
+Scope:
+Operator:
+Reviewer:
+
+Actions completed:
+- [ ] Account/session access stopped.
+- [ ] Roster updated.
+- [ ] KPI denominator/exclusion notes updated.
+- [ ] Export manifest updated or export blocked.
+- [ ] Deletion/anonymization completed where policy allows.
+
+Records not changed and reason:
+
+Reviewer decision:
+```
+
+## Daily Operator Checklist During Pilot
+
+- [ ] New sessions have consent receipts before start time.
+- [ ] No direct identifiers were added to evidence CSVs.
+- [ ] Instructor/admin access review entries are present for any manual review.
+- [ ] Withdrawal requests are processed before KPI/export generation.
+- [ ] Safety events are reviewed without copying sensitive free text into the roster.
+- [ ] Evidence root checker is run after each pilot day:
+
+```powershell
+python scripts\check-phase3-artifacts.py --check --evidence-root $root
+```
+
+## Exit Gate
+
+The privacy gate is complete only when:
+
+- `privacy_audit.md` is signed and has no unresolved high severity issue.
+- `withdrawal_log.csv` exists and all rows are terminal or explicitly marked
+  `needs_legal_review`.
+- KPI and export artifacts use the same withdrawal/exclusion decisions.
+- Legal/privacy reviewer has approved real participant evidence handling.
+