fix(gates): detect unmasked case pii

This commit is contained in:
2026-07-14 08:59:35 +08:00
parent 440ffb71ce
commit 8b81b5a382
3 changed files with 17 additions and 8 deletions
+2 -2
View File
@@ -48,8 +48,8 @@
### 2026-07-14 增量:BRD 内容数量门槛
- `scripts/demo-check.sh` 的严格模式现在额外检查已解析的住宅 SOP 文档数量 `>=5`(`aihr_knowledge_attach.type='sop' AND status=2`)和已入库案例数量 `>=20`(`aihr_case_record.status='已入库'`)。
- 这两个查询只证明数量,不证明 5 个核心流程覆盖、案例脱敏质量、每周每项目产出或视频学习内容;这些仍需业务内容复核和正式窗口数据。
- `scripts/demo-check.sh` 的严格模式现在额外检查已解析的住宅 SOP 文档数量 `>=5`(`aihr_knowledge_attach.type='sop' AND status=2`)、已入库案例数量 `>=20`(`aihr_case_record.status='已入库'`)和案例文本中已知手机号/邮箱/房号模式为 0。
- 这些查询只证明数量和已知模式脱敏,不证明 5 个核心流程覆盖、姓名等未结构化 PII、每周每项目产出或视频学习内容;这些仍需业务内容复核和正式窗口数据。
- `scripts/tests/demo-check-pilot-gates.test.sh` 已补充跨租户、错误状态和非 SOP 文档样本,防止严格门禁用错误数据刷过。
### P2/P3:明确暂缓,不应混入当前修复
+4 -2
View File
@@ -79,7 +79,7 @@ check_pilot_samples() {
[[ "$(pilot_sql_scalar "SELECT DATE('$pilot_end') IS NOT NULL")" == "1" ]] || fail "invalid AIHR_PILOT_END_DATE: $pilot_end"
[[ "$pilot_start" > "$pilot_end" ]] && fail "pilot start date must not be after end date"
local scenario_count completed_people pilot_people org_active org_phone_linked org_pilot_people org_pilot_projects org_completed_people audio_table_count prompt_version_schema_count sop_document_count case_stored_count
local scenario_count completed_people pilot_people org_active org_phone_linked org_pilot_people org_pilot_projects org_completed_people audio_table_count prompt_version_schema_count sop_document_count case_stored_count case_unmasked_count
local calibration_stats calibration_count calibration_matched sop_total sop_usable sop_pending satisfaction_count satisfaction_average
local org_identity_cte="
WITH active_org AS (
@@ -126,6 +126,7 @@ check_pilot_samples() {
prompt_version_schema_count="$(pilot_sql_scalar "SELECT COUNT(*) FROM information_schema.columns WHERE table_schema = DATABASE() AND table_name = 'aihr_sop_answer_review' AND column_name = 'prompt_version'")"
sop_document_count="$(pilot_sql_scalar "SELECT COUNT(*) FROM aihr_knowledge_attach WHERE tenant_id = '000000' AND type = 'sop' AND status = 2")"
case_stored_count="$(pilot_sql_scalar "SELECT COUNT(*) FROM aihr_case_record WHERE tenant_id = '000000' AND status = '已入库'")"
case_unmasked_count="$(pilot_sql_scalar "SELECT COUNT(*) FROM aihr_case_record WHERE tenant_id = '000000' AND status = '已入库' AND CONCAT_WS(' ', title, transcript, summary) REGEXP '1[3-9][0-9]{9}|[[:alnum:]._%+-]+@[[:alnum:].-]+\\.[[:alpha:]]{2,}|[0-9]{1,2}(栋|幢|号楼|单元)[0-9]{3,4}|[0-9]{1,2}[--][0-9]{3,4}'")"
completed_people="$(pilot_sql_scalar "
SELECT COUNT(*)
FROM (
@@ -267,12 +268,13 @@ check_pilot_samples() {
AND s.finished_time < DATE_ADD('${pilot_end}', INTERVAL 1 DAY)
)
SELECT COALESCE(ROUND(AVG(satisfaction_score), 1), 0) FROM formal_sessions WHERE satisfaction_score IS NOT NULL")"
echo "OK: pilot gates window=${pilot_start}..${pilot_end} scenarios=${scenario_count}/12 sop_docs=${sop_document_count}/5 stored_cases=${case_stored_count}/20 ten_sessions=${completed_people}/${pilot_people} org_login=${org_phone_linked}/${org_active} org_pilot=${org_pilot_people}/20 projects=${org_pilot_projects}/1-2 org_ten_sessions=${org_completed_people}/${org_pilot_people} calibration=${calibration_count}/20 consistency=${calibration_matched}/${calibration_count} sop_usable=${sop_usable}/${sop_total} sop_pending=${sop_pending} satisfaction=${satisfaction_average}/5 responses=${satisfaction_count} audio_schema=${audio_table_count} prompt_version_schema=${prompt_version_schema_count}"
echo "OK: pilot gates window=${pilot_start}..${pilot_end} scenarios=${scenario_count}/12 sop_docs=${sop_document_count}/5 stored_cases=${case_stored_count}/20 case_unmasked=${case_unmasked_count} ten_sessions=${completed_people}/${pilot_people} org_login=${org_phone_linked}/${org_active} org_pilot=${org_pilot_people}/20 projects=${org_pilot_projects}/1-2 org_ten_sessions=${org_completed_people}/${org_pilot_people} calibration=${calibration_count}/20 consistency=${calibration_matched}/${calibration_count} sop_usable=${sop_usable}/${sop_total} sop_pending=${sop_pending} satisfaction=${satisfaction_average}/5 responses=${satisfaction_count} audio_schema=${audio_table_count} prompt_version_schema=${prompt_version_schema_count}"
[[ "${AIHR_PILOT_STRICT:-false}" != "true" ]] || {
(( scenario_count >= 12 )) || fail "pilot enabled scenarios ${scenario_count}/12"
(( sop_document_count >= 5 )) || fail "pilot residential SOP documents ${sop_document_count}/5"
(( case_stored_count >= 20 )) || fail "pilot stored cases ${case_stored_count}/20"
(( case_unmasked_count == 0 )) || fail "pilot stored cases contain unmasked known PII (${case_unmasked_count})"
(( audio_table_count == 1 )) || fail "pilot audio schema unavailable (aihr_practice_audio missing)"
(( prompt_version_schema_count == 1 )) || fail "pilot RAG traceability schema unavailable (aihr_sop_answer_review.prompt_version missing)"
(( org_phone_linked >= 20 )) || fail "pilot mobile-login org members ${org_phone_linked}/20 (active org ${org_active})"
+11 -4
View File
@@ -57,7 +57,9 @@ CREATE DATABASE \`$TEST_DB\` CHARACTER SET utf8mb4 COLLATE utf8mb4_0900_ai_ci;
USE \`$TEST_DB\`;
CREATE TABLE aihr_practice_scenario (tenant_id varchar(20), position varchar(80), enabled tinyint);
CREATE TABLE aihr_knowledge_attach (tenant_id varchar(20), type varchar(50), status tinyint);
CREATE TABLE aihr_case_record (tenant_id varchar(20), status varchar(30));
CREATE TABLE aihr_case_record (
tenant_id varchar(20), status varchar(30), title varchar(120), transcript text, summary varchar(1000)
);
CREATE TABLE aihr_practice_session (tenant_id varchar(20), session_id varchar(100), ext_party_id varchar(100), mode varchar(30), finished_time datetime, satisfaction_score tinyint);
CREATE TABLE aihr_practice_assignment (tenant_id varchar(20), ext_party_id varchar(100), create_time datetime);
CREATE TABLE aihr_org_snapshot (
@@ -87,15 +89,15 @@ INSERT INTO aihr_knowledge_attach VALUES
('other','sop',2),
('000000','sop',1);
INSERT INTO aihr_case_record
SELECT '000000','已入库' FROM (
SELECT '000000','已入库', CONCAT('case-', n), '已按流程回访,手机号1390****1111', '已完成' FROM (
SELECT 1 n UNION ALL SELECT 2 UNION ALL SELECT 3 UNION ALL SELECT 4 UNION ALL SELECT 5
UNION ALL SELECT 6 UNION ALL SELECT 7 UNION ALL SELECT 8 UNION ALL SELECT 9 UNION ALL SELECT 10
UNION ALL SELECT 11 UNION ALL SELECT 12 UNION ALL SELECT 13 UNION ALL SELECT 14 UNION ALL SELECT 15
UNION ALL SELECT 16 UNION ALL SELECT 17 UNION ALL SELECT 18 UNION ALL SELECT 19 UNION ALL SELECT 20
) n;
INSERT INTO aihr_case_record VALUES
('000000','已转写'),
('other','已入库');
('000000','已转写','pending','',''),
('other','已入库','other','','');
INSERT INTO aihr_org_snapshot VALUES
('000000','active','A','P1','13900000001'),
('000000','active','B','P1','13900000001'),
@@ -145,6 +147,7 @@ pilot_output="$(AIHR_PILOT_START_DATE=2026-07-07 AIHR_PILOT_END_DATE=2026-07-10
[[ "$pilot_output" == *'org_ten_sessions=1/2'* ]] || { echo "FAIL: formal 9/10 boundary: $pilot_output" >&2; exit 1; }
[[ "$pilot_output" == *'calibration=1/20 consistency=1/1'* ]] || { echo "FAIL: calibration window: $pilot_output" >&2; exit 1; }
[[ "$pilot_output" == *'sop_docs=5/5 stored_cases=20/20'* ]] || { echo "FAIL: BRD content quantity gates: $pilot_output" >&2; exit 1; }
[[ "$pilot_output" == *'case_unmasked=0'* ]] || { echo "FAIL: redacted case content gate: $pilot_output" >&2; exit 1; }
[[ "$pilot_output" == *'sop_usable=1/1 sop_pending=1'* ]] || { echo "FAIL: SOP window: $pilot_output" >&2; exit 1; }
[[ "$pilot_output" == *'satisfaction=5.0/5 responses=19'* ]] || { echo "FAIL: safe satisfaction identity scope: $pilot_output" >&2; exit 1; }
@@ -153,6 +156,10 @@ pilot_output="$(AIHR_PILOT_START_DATE=2026-07-07 AIHR_PILOT_END_DATE=2026-07-10
[[ "$pilot_output" == *'org_ten_sessions=2/2'* ]] || { echo "FAIL: formal 10/10 boundary: $pilot_output" >&2; exit 1; }
[[ "$pilot_output" == *'satisfaction=5.0/5 responses=20'* ]] || { echo "FAIL: formal satisfaction scope after window update: $pilot_output" >&2; exit 1; }
pilot_sql_scalar "UPDATE aihr_case_record SET transcript = '手机号13900000000' WHERE tenant_id = '000000' AND status = '已入库' LIMIT 1" >/dev/null
pilot_output="$(AIHR_PILOT_START_DATE=2026-07-07 AIHR_PILOT_END_DATE=2026-07-10 check_pilot_samples)"
[[ "$pilot_output" == *'case_unmasked=1'* ]] || { echo "FAIL: unmasked case PII was not detected: $pilot_output" >&2; exit 1; }
if (AIHR_PILOT_STRICT=true AIHR_PILOT_START_DATE= AIHR_PILOT_END_DATE= check_pilot_samples) >/tmp/wygj-pilot-strict-test.log 2>&1; then
echo "FAIL: strict mode accepted an implicit date window" >&2
exit 1