参数化决策包
模块覆盖
| Module | Status | Score | Summary |
|---|---|---|---|
| identity_lock | complete | 90 | 锁定 App 身份、公司主体、商店 ID、官网和政策链接。 |
| market_proof | complete | 100 | 验证下载、评分、评论、榜单、第三方估算和 IAP/收入信号。 |
| review_mining | complete | 100 | 聚类订阅、首次结果、性能、隐私、客服退款等用户痛点。 |
| ad_library | partial | 66 | 追踪 TikTok、Meta、Google、DataEye/ADX 的素材、query 和 blocked 原因。 |
| store_screenshot_ocr | partial | 64 | 抽取商店截图中的 hook、before/after、价格和核心承诺。 |
| ad_product_consistency | missing | 12 | 对照广告素材承诺和 App 内第一个真实结果，判断是否能兑现。 |
| paywall_pricing_trust | partial | 65 | 拆 paywall moment、试用、订阅周期、续费价、退款和取消路径。 |
| privacy_data_safety | complete | 93 | 对照隐私政策、Data Safety、权限、第三方处理、删除和留存。 |
| support_refund_quality | partial | 63 | 检查 support 入口、退款路径、删除请求、客服和平台退款归属。 |
| trust_score | partial | 65 | 评估隐私、退款、support、审计、授权、可取消和可撤销等信任要素。 |
| first_result_score | partial | 63 | 判断首次 30 秒结果是否清晰、可感知、可截图、足以触发继续或付费。 |
| creative_replicability | partial | 63 | 判断 1 周内能否生产 10-50 条素材，以及 hook 是否能复用。 |
| supply_difficulty | missing | 12 | 拆内容、模型、权限、数据源、人工 QA、服务器和运营瓶颈。 |
| refund_risk | partial | 85 | 识别退款、取消、客服、试用和续费争议，估算客服/品牌风险。 |
| mvp_gate | complete | 94 | 明确 1 周、1 个月、3 个月 scope，写清 kill gate 和必须排除项。 |
| safe_wedge | complete | 96 | 拆可做安全 wedge 和不能复制的反模式，防止把灰色需求误读成机会。 |
| dark_pattern_score | partial | 84 | 识别误导免费、隐藏续费、取消困难、先付费后结果等反模式。 |
| anti_pattern_library | partial | 64 | 沉淀不能复制的模式，供后续竞品和开发前 guardrail 复用。 |
| run_retro | complete | 90 | 每轮结束回写流程改进、blocked source、stale cleanup 和下轮动作。 |
突破点和好做增强
下一轮突破点
- Google Play URL | 未验证到官方 Google Play 包 | blocked
- 1 周 MVP：本地/私有库内本人照片重复和相似素材整理，支持删除/导出/授权备注。
- 1 个月 MVP：授权素材库、相似图检索、使用台账、撤销授权、审计日志。
- 设计本人/授权素材管理 MVP guardrail | validation-analyst | manual | 2026-06-16
好做且高价值
- 当前判断：`kill_core_feature`，只推进本人/授权照片整理、肖像授权素材管理、反盗图/公开曝光自查等安全 wedge。
- 创作者/演员/模特授权素材库和肖像使用台账。
- 反盗图/反冒用，但只服务本人或权利方，输出“素材出处/相似图片/授权状态”，不输出陌生人身份。
- 广告素材、短剧角色和 AI 生成资产管理，不识别真实身份。
- 1 周 MVP：本地/私有库内本人照片重复和相似素材整理，支持删除/导出/授权备注。
- 1 个月 MVP：授权素材库、相似图检索、使用台账、撤销授权、审计日志。
- 技术边界：默认本地或私有库内 1:1 / 授权 1:N；结果是“素材相似/授权状态/来源证据”，不是“这个人是谁”。
- 设计本人/授权素材管理 MVP guardrail | validation-analyst | manual | 2026-06-16
不能复制的反模式
- Verdict：kill_core_feature / promote_safe_wedge
- 一句话：Sherlock 的核心“陌生人照片 -> 身份/社媒/犯罪线索”能力应 kill；只能学习需求，不建议复刻。
- 当前判断：`kill_core_feature`，只推进本人/授权照片整理、肖像授权素材管理、反盗图/公开曝光自查等安全 wedge。
- 反 catfish / 约会安全 | App Store 和官网定位 | 需求强，但极易滑向陌生人身份搜索
- 滥用用途 | 评论中出现查亲友、邻居、troll、scary 等 | 产品必须拒绝陌生人识别和跟踪场景
- 生物识别 | GDPR Art. 9 将用于唯一识别自然人的 biometric data 列为特殊类别 | 阻断陌生人搜索
- 反盗图/反冒用，但只服务本人或权利方，输出“素材出处/相似图片/授权状态”，不输出陌生人身份。
四大类执行队列
| Category | Priority | Item | Status | Owner | Next action |
|---|---|---|---|---|---|
| core_structure | P0 | 身份锁定 | done | validation-analyst | 补齐缺失平台 ID |
| core_structure | P0 | JTBD / 首次价值 | pending | app-dissector | 明确用户触发时刻、替代方案、首次 30 秒结果和愿付费原因。 |
| core_structure | P0 | Paywall / Pricing / Trust | pending | validation-analyst | 补 SKU / price matrix |
| core_structure | P0 | MVP Gate | done | validation-analyst | 把 promote 候选转成 MVP spec |
| high_value_collection | P0 | 评论聚类 | done | validation-analyst | 采集最近 7/30/90 天评论 |
| high_value_collection | P0 | 商店截图 OCR | pending | creative-analyst | 补商店截图 OCR |
| high_value_collection | P0 | 广告库查询 | pending | creative-analyst | 跑 ad-library-queries.csv |
| high_value_collection | P1 | IAP / Pricing Matrix | pending | validation-analyst | 记录 SKU、trial、billed upfront、renewal price、refund route。 |
| high_value_collection | P1 | 隐私 / Data Safety 对照 | done | compliance-reviewer | 补 privacy-vs-data-safety 表 |
| breakthrough_direction | P0 | 广告承诺与产品交付一致性 | pending | app-dissector | 补 top creative -> first result 对照表 |
| breakthrough_direction | P0 | 安全可做 Wedge | done | compliance-reviewer | 沉淀 anti-pattern library |
| breakthrough_direction | P1 | 单位经济账 | pending | validation-analyst | 估算 CPI、试用率、订阅率、退款率、模型/内容/客服成本。 |
| breakthrough_direction | P1 | 供给难度拆解 | pending | vendor-integrator | 补 supply/cost table |
| easy_high_priority_enhancement | P0 | HTML 决策包 | done | validation-analyst | 让每个 research 都有可读、可比较、可决策的阅读版。 |
| easy_high_priority_enhancement | P0 | Research Queue | done | validation-analyst | 把核心结构、高价值采集、突破方向、好做增强排队执行。 |
| easy_high_priority_enhancement | P1 | 反模式库 | pending | compliance-reviewer | 把本机会反模式写入 research-retro.md |
| easy_high_priority_enhancement | P1 | Run Retro / Backlog 回写 | done | validation-analyst | 更新 research-retro.md |
真实采集面板
Pricing / Paywall Matrix
| sku | product_type | period | trial | price | currency | renewal_price | refund_route | status |
|---|---|---|---|---|---|---|---|---|
| app_store:US:install | free_install | 0 | USD | App Store refund / subscription management; in-app cancellation path not captured | partial | |||
| app_store:GB:install | free_install | 0 | GBP | App Store refund / subscription management; in-app cancellation path not captured | partial |
Store Screenshot OCR / VL
| platform | country | hook | promise | status | next_action |
|---|---|---|---|---|---|
| blocked | store detail 未返回截图，补浏览器截图或记录 blocked reason。 |
Reddit Relevance
| query | subreddit | title | relevance_score | relevance_status | status |
|---|---|---|---|---|---|
| Sherlock AI Face Search | PromoCodeShare | Sherlock Ai Face Search | 100 | high_confidence_target | partial |
| Sherlock AI Face Search | PromoCodeShare | Sherlock AI code | 97 | high_confidence_target | partial |
| Sherlock AI Face Search | PromoCodeShare | Sherlock ai face search | 100 | high_confidence_target | partial |
| Sherlock AI Face Search | PromoCodeShare | Sherlock Ai: Face search 6-digit Code | 100 | high_confidence_target | partial |
| Sherlock AI Face Search | PromoCodeShare | Sherlock: AI Face Search App | 100 | high_confidence_target | partial |
| Sherlock AI Face Search | SwipeHelper | Has anyone used AI facial recognition to find someone on a dating app? | 8 | low_relevance_archive | needs_archive |
| Sherlock AI Face Search | PromoCodeShare | Sherlock : AI Face Search Refer Code | 100 | high_confidence_target | partial |
| Sherlock AI Face Search | CAIRevolution | My opnion in C.Ai alternatives (for those who are searching for those after the new shitty updates) + "little" rant at the end | 8 | low_relevance_archive | needs_archive |
显示 8 / 25 行。
Privacy / Support / Terms
| data_category | third_party_processor | deletion_path | risk_level | status | source_url |
|---|---|---|---|---|---|
| trust_url | high | verified | https://www.imsherlock.com/ | ||
| privacy_policy | otice? How can you contact us about this notice? How can you review, update, or delete the data we collect from you? What information do we collect? Personal information you disclose to us In Short: We collect personal… | low | verified | https://www.imsherlock.com/privacy | |
| privacy_policy | p’s privacy policy and required by these Guidelines. Explain its data retention/deletion policies and describe how a user can revoke consent and/or request deletion of the user’s data. (ii) Permission: Apps that collect… | low | verified | https://developer.apple.com/app-store/review/guidelines/ | |
| support_contact | thin your app, then it must also allow users to request for their account to be deleted. Users must have a readily discoverable option to initiate app account deletion from within your app and outside of your app (for e… | low | verified | https://support.google.com/googleplay/android-developer/answer/10144311?hl=en |
Support / Refund Quality
| signal_type | claim | status | risk_level |
|---|---|---|---|
| support_contact | support email: support@imsherlock.com | partial | medium |
| refund_policy | f items that are eligible for in-app purchase to others. Such gifts may only be refunded to the original purchaser and may not be exchanged. Apps distributed via the Mac App Store may host plug-ins or extensions that ar… | partial | medium |
| cancel_path | mation: The length of the renewal term and the fact that it will continue until canceled What will be provided during each period The actual charges that will be billed to the customer How to cancel 4.10 Monetizing Buil… | partial | medium |
| delete_request | otice? How can you contact us about this notice? How can you review, update, or delete the data we collect from you? What information do we collect? Personal information you disclose to us In Short: We collect personal… | partial | medium |
Ad Library Query Status
| platform_label | query | status | purpose |
|---|---|---|---|
| TikTok Creative Center | Sherlock: AI Face Search | pending | collect creative URLs, screenshots, hook, landing page, country, blocked reason |
| TikTok Creative Center | Sherlock | pending | collect creative URLs, screenshots, hook, landing page, country, blocked reason |
| Meta Ad Library | Sherlock: AI Face Search | pending | collect creative URLs, screenshots, hook, landing page, country, blocked reason |
| Meta Ad Library | Sherlock | pending | collect creative URLs, screenshots, hook, landing page, country, blocked reason |
| Google Ads Transparency | Sherlock: AI Face Search | pending | collect creative URLs, screenshots, hook, landing page, country, blocked reason |
| Google Ads Transparency | Sherlock | pending | collect creative URLs, screenshots, hook, landing page, country, blocked reason |
| DataEye / ADX | Sherlock: AI Face Search | blocked | No DataEye creative match in current inputs; needs product detail or external ad library. |
| DataEye / ADX | Sherlock | blocked | No DataEye creative match in current inputs; needs product detail or external ad library. |
Ad Creative Signals
| match_type | product_title | creative_title | media | rank | exposure | play | status | original_url | video_url |
|---|---|---|---|---|---|---|---|---|---|
| no_dataeye_creative_match | Sherlock: AI Face Search | blocked |
最新 Research Queue
| category | priority | item_id | status | owner | next_action |
|---|---|---|---|---|---|
| core_structure | P0 | identity_lock | done | validation-analyst | 补齐缺失平台 ID |
| core_structure | P0 | jtbd_first_value | pending | app-dissector | 明确用户触发时刻、替代方案、首次 30 秒结果和愿付费原因。 |
| core_structure | P0 | paywall_pricing_trust | pending | validation-analyst | 补 SKU / price matrix |
| core_structure | P0 | mvp_gate | done | validation-analyst | 把 promote 候选转成 MVP spec |
| high_value_collection | P0 | review_mining | done | validation-analyst | 已真实采集 60 条商店评论；下一步补 Reddit/Trustpilot 和人工复核。 |
| high_value_collection | P0 | store_screenshot_ocr | blocked | creative-analyst | 商店截图 OCR 未形成可用文本；processed=0, empty=0, failed=0, skipped=1。补浏览器截图或 VL fallback。 |
| high_value_collection | P0 | ad_library | pending | creative-analyst | 跑 ad-library-queries.csv |
| high_value_collection | P1 | iap_pricing_matrix | partial | validation-analyst | 已扩展到 2 条价格/退款/取消文本线索；下一步真机 SKU/paywall capture。 |
| high_value_collection | P1 | privacy_data_safety | done | compliance-reviewer | 已解析 4 个 privacy/support/terms 页面；下一步人工核对 Data Safety 和 App 内删除入口。 |
| breakthrough_direction | P0 | ad_product_consistency | pending | app-dissector | 补 top creative -> first result 对照表 |
| breakthrough_direction | P0 | safe_wedge | done | compliance-reviewer | 沉淀 anti-pattern library |
| breakthrough_direction | P1 | unit_economics | pending | validation-analyst | 估算 CPI、试用率、订阅率、退款率、模型/内容/客服成本。 |
| breakthrough_direction | P1 | supply_difficulty | pending | vendor-integrator | 补 supply/cost table |
| easy_high_priority_enhancement | P0 | decision_pack_html | done | validation-analyst | 让每个 research 都有可读、可比较、可决策的阅读版。 |
| easy_high_priority_enhancement | P0 | research_queue | done | validation-analyst | 把核心结构、高价值采集、突破方向、好做增强排队执行。 |
| easy_high_priority_enhancement | P1 | anti_pattern_library | pending | compliance-reviewer | 把本机会反模式写入 research-retro.md |
| easy_high_priority_enhancement | P1 | run_retro | done | validation-analyst | 更新 research-retro.md |
摘要
- 一句话：Sherlock 的核心“陌生人照片 -> 身份/社媒/犯罪线索”能力应 kill；只能学习需求，不建议复刻。
- 为什么进入深挖：AI face search 有强付费焦虑，但同时触发生物识别、隐私、跟踪骚扰和平台政策红线。
- 当前判断：`kill_core_feature`，只推进本人/授权照片整理、肖像授权素材管理、反盗图/公开曝光自查等安全 wedge。
关键发现
| 维度 | 结论 | 状态 |
|---|---|---|
| Product promise | 用一张人脸照片搜索某人的公开在线存在，包括社媒资料、博客、视频、新闻页面；主场景是 dating safety / 反 catfish | verified |
| Sensitive output | 输出相似/匹配人脸结果和公开网页线索，涉及身份推断和潜在犯罪/sex offender 风险线索 | verified |
| Processor | 隐私条款称照片会传给第三方 image search service `Tech Solutions - Belize`，Sherlock 自称不存储照片 | verified |
| Reviews | App Store RSS 320 条评论样本：付费和准确率是主要投诉；存在越界用途自报 | verified |
| Compliance | GDPR Art. 9、CCPA/CPRA、BIPA、Apple/Google policy、FTC 都构成高风险边界 | verified |
用户研究
| Cluster | 证据 | 机会 / 风险 |
|---|---|---|
| 反 catfish / 约会安全 | App Store 和官网定位 | 需求强，但极易滑向陌生人身份搜索 |
| 准确率差 | 320 条评论中 97 条触发不准/随机/找不到等关键词 | 误识别会把无辜第三方卷入身份指认 |
| 付费 / credits / refund | 151 条触发付费/订阅/coins/refund 关键词 | 先付费后无结果会造成强退款风险 |
| 隐私和第三方处理 | 隐私页披露第三方搜索服务 | 上传第三方人脸和公开索引拼接是主要风险 |
| 滥用用途 | 评论中出现查亲友、邻居、troll、scary 等 | 产品必须拒绝陌生人识别和跟踪场景 |
合规红线
| 风险 | 结论 | 是否阻断 |
|---|---|---|
| 生物识别 | GDPR Art. 9 将用于唯一识别自然人的 biometric data 列为特殊类别 | 阻断陌生人搜索 |
| BIPA / CCPA | face geometry / facial recognition 属于敏感个人信息范围 | 阻断公开人脸索引 |
| Apple policy | 不允许在无明确同意下从公开数据库编译个人信息 | 阻断身份聚合 |
| Google Play policy | 敏感数据需要显著披露、明确同意、限制用途和安全处理 | 阻断低披露处理 |
| 滥用和安全 | 跟踪、骚扰、未成年人、约会对象、犯罪断言都是高风险场景 | 阻断 |
不建议做
- 公开互联网人脸索引。
- 陌生人身份识别。
- 约会对象背景搜索。
- 按脸找社媒。
- 犯罪 / sex offender 线索聚合。
- 绕平台隐私的相似脸搜索。
可推进的安全 wedge
- 本人照片互联网曝光自查。
- 授权头像库相似/重复检测。
- 创作者/演员/模特授权素材库和肖像使用台账。
- 反盗图/反冒用，但只服务本人或权利方，输出“素材出处/相似图片/授权状态”，不输出陌生人身份。
- 广告素材、短剧角色和 AI 生成资产管理，不识别真实身份。
可做性
- 1 周 MVP：本地/私有库内本人照片重复和相似素材整理，支持删除/导出/授权备注。
- 1 个月 MVP：授权素材库、相似图检索、使用台账、撤销授权、审计日志。
- 3 个月版本：权利方反盗图公开曝光自查，但必须只处理本人/授权主体并做 abuse review。
- 技术边界：默认本地或私有库内 1:1 / 授权 1:N；结果是“素材相似/授权状态/来源证据”，不是“这个人是谁”。
判断
- Score：core feature 20 / safe wedge 72
- Verdict：kill_core_feature / promote_safe_wedge
- 理由：市场需求真实，但核心能力合规和滥用风险不可接受；安全 wedge 可借需求，不借功能边界。
下一步
| 动作 | Owner | Mode | Due |
|---|---|---|---|
| 写 `core_feature_kill vs safe_wedge_promote` 模板字段 | compliance-reviewer | manual | 2026-06-16 |
| 设计本人/授权素材管理 MVP guardrail | validation-analyst | manual | 2026-06-16 |
| 把人脸搜索类纳入高风险 kill gate | compliance-reviewer | manual | 2026-06-16 |
本轮流程改进记录
| 问题 | 影响 | 建议 | 状态 |
|---|---|---|---|
| 需求强不等于可复刻 | 容易误把灰色需求当机会 | 模板强制区分 core_feature_kill 和 safe_wedge_promote | backlog_open |
| APK/SEO 站点容易混入官方 identity | 可能错误判断 Android 版本 | 单列 unverified_android_sources | backlog_open |
| face-search 类缺少固定合规字段 | 报告容易漏生物识别处理链路 | 增加 biometric_risk_class 和 third_party_processor 字段 | backlog_open |
证据账本
| 类型 | 来源 | 结论 | 状态 | 置信度 |
|---|---|---|---|---|
| manual_request | user_requested_batch | 用户指定进入本轮机会深挖队列。 | verified | high |
| store | Apple App Store | iOS App Store 显示 Sherlock: AI Face Search，开发者 Zachary Yudenfriend，主承诺是人脸照片搜索在线存在。 | verified | high |
| api | Apple iTunes Lookup | Apple Lookup 返回 bundleId com.zachyudenfriend.sherlockapp。 | verified | high |
| official_site | Sherlock official site | 官网定位 dating safety by photo lookup，页脚显示 BzH App Studio LLC。 | verified | high |
| privacy | Sherlock Privacy | 隐私页称搜索照片传给 Tech Solutions - Belize 第三方 image search service，Sherlock 自称不存储照片。 | verified | medium |
| reviews | App Store RSS | 公开 RSS 320 条评论样本显示付费/credits/refund 和准确率是主要投诉，并存在越界用途自报。 | verified | medium |
| policy | Apple Review Guidelines | Apple policy 对无明确同意编译公开数据库个人信息构成阻断风险。 | verified | high |
| law | GDPR Article 9 | GDPR Article 9 将用于唯一识别自然人的 biometric data 列为特殊类别。 | verified | high |
| law | Illinois BIPA | BIPA 将 face geometry scan 和派生识别信息列入 biometric categories。 | verified | high |
| policy | Google Play User Data | Google Play 对敏感用户数据要求披露、同意、安全处理和限制用途。 | verified | high |
| ads | Meta/TikTok public ad libraries | 本轮未验证直接 active ad creative；评论提到 Instagram/TikTok/Reels ads。 | blocked | low |
| store | Google Play | 未验证到官方 Google Play package；APK 站点不视为官方 identity。 | blocked | medium |
| store_detail_auto | App Store / Google Play collector | 真实采集 2 个商店详情，包含评分、评论量、版本、截图和隐私链接字段；截图 URL 0 个，隐私链接 0 个。 | verified | high |
| review_sample_auto | Store review collector | 真实采集 60 条商店评论样本，platforms=app_store，已回填 review-clusters.csv。 | verified | high |
| pricing_auto | Store listing pricing fields | 真实回填 2 条商店价格/IAP range 信号到 pricing-matrix.csv；SKU、trial、续费价和 App 内取消路径仍需深采。 | verified | high |
| community_signal_auto | Reddit search collector | 真实采集 Reddit search leads 25 条；当前为待 relevance review 的社区线索，不直接当作已验证产品评论。Top: Sherlock Ai Face Search; Sherlock AI code; Sherlock ai face search | partial | medium |
| store_screenshot_ocr_auto | Store screenshot OCR | 真实处理 0 张商店截图 OCR：complete=0, empty=0, failed=0, skipped=1；complete 包含 OCR 增强/VL fallback；输出见 assets/store-screenshot-ocr.csv。 | partial | medium |
| community_relevance_auto | Reddit relevance scorer | 已对 25 条 Reddit leads 做相关性门禁：high=14, review=0, low_archive=11。Top review candidates: Sherlock Ai Face Search; Sherlock AI code; Sherlock ai face search; Sherlock Ai: Face search 6-digit Code; Sherlock: AI Face Search App | verified | high |
| trust_pricing_resolver_auto | Privacy/support/terms/pricing resolver | 真实解析 trust/pricing URL 4 个：fetch_success=4, blocked=0, privacy_rows=4, support_rows=4, pricing_rows=2。 | verified | high |
| ad_creative_digest_auto | DataEye creative digest | 已生成 DataEye creative digest：exact_or_related=0, lane_benchmark=0。Top: | partial | medium |