name: vkc-moderation-trust
description: AI moderation, trust scoring, answer-accuracy, and keyword flag queue workflow for VKC UGC. Use when changing content-filter, moderation queues, trust badges, or answer-accuracy review. (키워드= moderation, trust, keywordFlagQueue, answer-accuracy, content-filter, WO-009)
metadata:
short-description: Moderation + trust + answer accuracy
VKC Moderation & Trust (stub)
Always pair with $vkc-repo-guardrails. Read audit SoT before mutation.
When to use
- Pre/post moderation pipelines (
content-filter, report → auto-hide)
- Answer accuracy sampling and admin review routes
- Trust score, adoption rewards, badge eligibility
- WO-009 AI moderation safety contract implementation
Canonical docs
| Doc |
Purpose |
docs/strategy/AI_MODERATION_TRUST_AUDIT_2026-05-24.md |
Gap matrix, ~37% complete vs guideline |
docs/ops/ANSWER_ACCURACY_BOT_SPEC_2026-05-24.md |
Regex + review queue spec |
docs/ops/ai_moderation_safety_contract.md |
Enum actions before OpenAI |
docs/engineering/WORK_ORDERS_DEV_2026-05-24.md |
WO-009 (AI classifier), WO-014 (writeGuard), WO-015 (admin AI labels) |
Key code paths
| Area |
Path |
| Submit-time filter |
src/lib/content-filter.ts |
| Keyword flag queue |
src/lib/moderation/keywordFlagQueue.ts |
| Report → auto-hide |
src/lib/moderation/contentModeration.ts |
| Answer accuracy patterns |
config/ops/answer-accuracy/regex-patterns.json |
| Admin review API |
src/app/api/admin/ops/answer-accuracy/review/route.ts |
| Trust presentation |
src/lib/utils/trustBadges.ts, src/lib/community/reputation.ts |
| Sample script |
scripts/ops/answer-accuracy-sample.mjs |
Commands
node scripts/ops/answer-accuracy-sample.mjs
npm run test:unit -- --test-name-pattern moderation
E2E_PORT=3100 npx playwright test e2e/moderation-auto-hide.spec.ts --project=chromium
Non-negotiables
- Follow
docs/ops/ai_moderation_safety_contract.md before wiring any LLM classifier (WO-009).
- Do not log passport, ARC, phone, or raw verification evidence.
- Prefer publish-first for links/contacts; pre-block only hate/spam/illegal (audit P1 gap — align with product policy).
Lane map
- Roster owner:
interactive:worker-moderation
- Skill router:
config/ops/cursor/skill-router.json → worker-moderation
By