# goldenmatch-monorepo - Doramagic AI Context Pack

> 定位：安装前体验与判断资产。它帮助宿主 AI 有一个好的开始，但不代表已经安装、执行或验证目标项目。

## 充分原则

- **充分原则，不是压缩原则**：AI Context Pack 应该充分到让宿主 AI 在开工前理解项目价值、能力边界、使用入口、风险和证据来源；它可以分层组织，但不以最短摘要为目标。
- **压缩策略**：只压缩噪声和重复内容，不压缩会影响判断和开工质量的上下文。

## 给宿主 AI 的使用方式

你正在读取 Doramagic 为 goldenmatch-monorepo 编译的 AI Context Pack。请把它当作开工前上下文：帮助用户理解适合谁、能做什么、如何开始、哪些必须安装后验证、风险在哪里。不要声称你已经安装、运行或执行了目标项目。

## Claim 消费规则

- **事实来源**：Repo Evidence + Claim/Evidence Graph；Human Wiki 只提供显著性、术语和叙事结构。
- **事实最低状态**：`supported`
- `supported`：可以作为项目事实使用，但回答中必须引用 claim_id 和证据路径。
- `weak`：只能作为低置信度线索，必须要求用户继续核实。
- `inferred`：只能用于风险提示或待确认问题，不能包装成项目事实。
- `unverified`：不得作为事实使用，应明确说证据不足。
- `contradicted`：必须展示冲突来源，不得替用户强行选择一个版本。

## 它最适合谁

- **正在使用 Claude/Codex/Cursor/Gemini 等宿主 AI 的开发者**：README 或插件配置提到多个宿主 AI。 证据：`README.md` Claim：`clm_0020` supported 0.86

## 它能做什么

- **Entity Resolution / Deduplication**（可做安装前预览）：Core GoldenMatch deduplication engine with fuzzy matching, blocking strategies, and probabilistic Fellegi-Sunter model for identifying duplicate records. 证据：`packages/python/goldenmatch/README.md`, `README.md`, `docs/adr/README.md` Claim：`clm_0001` supported 0.86, `clm_0003` supported 0.86, `clm_0008` supported 0.86, `clm_0009` supported 0.86 等
- **Data Validation / Profiling**（可做安装前预览）：GoldenCheck discovers validation rules from data automatically, profiling 10+ column characteristics and cross-column relationships without manual rule authoring. 证据：`packages/python/goldencheck/README.md`, `packages/python/goldencheck/CLAUDE.md`, `packages/python/goldencheck/goldencheck/profilers/CLAUDE.md` Claim：`clm_0002` supported 0.86, `clm_0006` supported 0.86, `clm_0007` supported 0.86, `clm_0012` supported 0.86 等
- **Data Standardization / Normalization**（可做安装前预览）：GoldenFlow standardizes data with domain-specific transform packs (healthcare, finance, ecommerce) and confidence scoring for each transformation. 证据：`README.md`, `examples/sql/README.md` Claim：`clm_0001` supported 0.86, `clm_0003` supported 0.86, `clm_0005` supported 0.86, `clm_0008` supported 0.86 等
- **Privacy-Preserving Record Linkage (PPRL)**（需要安装后验证）：Two-party privacy-preserving entity resolution using Bloom filter encoding — raw PII never crosses organizational boundaries. 证据：`examples/python/README.md`, `examples/airflow/README.md`, `examples/python/04_pprl_two_party.py` Claim：`clm_0004` supported 0.86, `clm_0008` supported 0.86, `clm_0010` supported 0.86, `clm_0013` supported 0.86 等
- **SQL UDFs (DuckDB & PostgreSQL)**（需要安装后验证）：Native SQL interface to GoldenMatch core APIs and GoldenFlow transforms, registered as UDFs callable directly in SQL queries. 证据：`examples/sql/README.md`, `examples/sql/duckdb_core_apis.sql`, `packages/rust/extensions/duckdb/README.md` Claim：`clm_0003` supported 0.86, `clm_0005` supported 0.86
- **CLI Commands**（可做安装前预览）：Rich CLI with 11+ commands for dedupe, scan, validate, review, diff, watch, fix, learn, baseline, and MCP serving. 证据：`packages/python/goldencheck/CLAUDE.md`, `examples/typescript/README.md` Claim：`clm_0002` supported 0.86, `clm_0006` supported 0.86, `clm_0007` supported 0.86, `clm_0011` supported 0.86 等
- **Interactive TUI**（可做安装前预览）：Textual-based terminal UI for interactive data scanning, validation review, and finding exploration. 证据：`packages/python/goldencheck/README.md`, `packages/python/goldencheck/CLAUDE.md` Claim：`clm_0002` supported 0.86, `clm_0006` supported 0.86, `clm_0007` supported 0.86, `clm_0012` supported 0.86 等
- **Web UI Workbench**（需要安装后验证）：FastAPI + React web interface for rule editing, pair comparison, sensitivity sweeps, and run history analysis. 证据：`README.md`, `examples/python/README.md` Claim：`clm_0001` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86, `clm_0008` supported 0.86 等
- **MCP Server**（需要安装后验证）：Model Context Protocol server exposing suite tools as AI-agent callable tools, hosted on Railway and registered on Smithery. 证据：`README.md`, `CLAUDE.md`, `packages/python/goldencheck/goldencheck/mcp/CLAUDE.md`, `examples/python/06_mcp_client.py` Claim：`clm_0001` supported 0.86, `clm_0003` supported 0.86, `clm_0008` supported 0.86, `clm_0009` supported 0.86 等
- **Airflow Integration**（需要安装后验证）：12 production-shaped Airflow DAGs for daily/incremental/warehouse-native dedupe, customer 360, PPRL, quality gates, and review workers. 证据：`examples/airflow/README.md`, `examples/airflow/golden_suite_identity_graph.py` Claim：`clm_0004` supported 0.86, `clm_0010` supported 0.86, `clm_0013` supported 0.86, `clm_0014` supported 0.86 等
- **TypeScript / Edge Runtime Support**（可做安装前预览）：Full TypeScript port with edge-safe core (/core) and Node-specific paths (/node), strict type checking, and npm distribution. 证据：`examples/typescript/README.md`, `examples/typescript/02-edge-runtime.ts` Claim：`clm_0006` supported 0.86, `clm_0011` supported 0.86
- **LLM Enhancement**（需要安装后验证）：AI-powered validation rule generation and finding explanation via Anthropic/OpenAI, with budget tracking and confidence scoring. 证据：`packages/python/goldencheck/goldencheck/llm/CLAUDE.md`, `packages/python/goldencheck/CLAUDE.md` Claim：`clm_0002` supported 0.86, `clm_0006` supported 0.86, `clm_0007` supported 0.86, `clm_0012` supported 0.86 等
- **Identity Graph**（可做安装前预览）：Stable entity_id assignment across runs with conflict detection, manual merge/split, and full event-log audit trail. 证据：`examples/python/README.md`, `examples/airflow/README.md` Claim：`clm_0004` supported 0.86, `clm_0008` supported 0.86, `clm_0010` supported 0.86, `clm_0013` supported 0.86 等
- **Review Workflow / Active Learning**（可做安装前预览）：Human-in-the-loop review queue for borderline pairs with Learning Memory feedback loop that retrains boost classifier from labeled examples. 证据：`examples/python/README.md`, `examples/airflow/README.md` Claim：`clm_0004` supported 0.86, `clm_0008` supported 0.86, `clm_0010` supported 0.86, `clm_0013` supported 0.86 等
- **Schema Mapping (InferMap)**（可做安装前预览）：Automatic schema alignment across heterogeneous data sources using InferMap, enabling multi-source customer unification. 证据：`examples/python/README.md`, `examples/airflow/README.md` Claim：`clm_0004` supported 0.86, `clm_0008` supported 0.86, `clm_0010` supported 0.86, `clm_0013` supported 0.86 等
- **Baseline / Drift Detection**（可做安装前预览）：Statistical baseline profiling with 13 check types for detecting distribution drift, schema changes, and data quality regressions. 证据：`packages/python/goldencheck/CLAUDE.md`, `packages/python/goldencheck/goldencheck/profilers/CLAUDE.md` Claim：`clm_0002` supported 0.86, `clm_0006` supported 0.86, `clm_0007` supported 0.86, `clm_0012` supported 0.86 等
- **Domain Packs**（可做安装前预览）：Pre-built validation rules for healthcare, finance, and ecommerce domains with industry-specific type definitions and checks. 证据：`packages/python/goldencheck-types/CLAUDE.md`, `packages/python/goldencheck/README.md`, `packages/python/goldencheck/goldencheck/mcp/CLAUDE.md` Claim：`clm_0002` supported 0.86, `clm_0007` supported 0.86, `clm_0009` supported 0.86, `clm_0017` supported 0.86
- **GitHub Actions**（需要安装后验证）：CI/CD integration via GitHub Actions for GoldenCheck, GoldenFlow, and GoldenMatch with PR comments and pass/fail status checks. 证据：`packages/actions/goldencheck/README.md`, `packages/actions/goldenflow/README.md`, `packages/actions/goldenmatch/README.md` Claim：`clm_0018` supported 0.86
- **Auto-Configuration (Autoconfig)**（可做安装前预览）：Zero-config deduplication that automatically selects optimal blocking strategy, match keys, and thresholds based on data characteristics. 证据：`docs/adr/README.md`, `docs/adr/2026-05-21-v1-13-autoconfig-roadmap.md` Claim：`clm_0001` supported 0.86, `clm_0019` supported 0.86

## 怎么开始

- `pip install goldenmatch && goldenmatch dedupe customers.csv` 证据：`README.md` Claim：`clm_0021` supported 0.86
- `npm install goldenmatch` 证据：`README.md` Claim：`clm_0022` supported 0.86
- `pip install 'goldenmatch[web]'` 证据：`README.md` Claim：`clm_0023` supported 0.86, `clm_0038` supported 0.86
- `pip install goldenmatch                    # core (CSV in, CSV out)` 证据：`README.md` Claim：`clm_0024` supported 0.86
- `pip install goldenmatch[embeddings]        # + sentence-transformers, FAISS` 证据：`README.md` Claim：`clm_0025` supported 0.86
- `pip install goldenmatch[llm]               # + Claude / OpenAI for LLM boost` 证据：`README.md` Claim：`clm_0026` supported 0.86
- `pip install goldenmatch[postgres]          # + Postgres sync` 证据：`README.md` Claim：`clm_0027` supported 0.86
- `pip install goldenmatch[snowflake]         # + Snowflake connector` 证据：`README.md` Claim：`clm_0028` supported 0.86
- `pip install goldenmatch[bigquery]          # + BigQuery connector` 证据：`README.md` Claim：`clm_0029` supported 0.86
- `pip install goldenmatch[databricks]        # + Databricks connector` 证据：`README.md` Claim：`clm_0030` supported 0.86

## 继续前判断卡

- **当前建议**：需要管理员/安全审批
- **为什么**：继续前可能涉及密钥、账号、外部服务或敏感上下文，建议先经过管理员或安全审批。

### 30 秒判断

- **现在怎么做**：需要管理员/安全审批
- **最小安全下一步**：先跑 Prompt Preview；若涉及凭证或企业环境，先审批再试装
- **先别相信**：角色质量和任务匹配不能直接相信。
- **继续会触碰**：角色选择偏差、命令执行、宿主 AI 配置

### 现在可以相信

- **适合人群线索：正在使用 Claude/Codex/Cursor/Gemini 等宿主 AI 的开发者**（supported）：有 supported claim 或项目证据支撑，但仍不等于真实安装效果。 证据：`README.md` Claim：`clm_0020` supported 0.86
- **能力存在：Entity Resolution / Deduplication**（supported）：可以相信项目包含这类能力线索；是否适合你的具体任务仍要试用或安装后验证。 证据：`packages/python/goldenmatch/README.md`, `README.md`, `docs/adr/README.md` Claim：`clm_0001` supported 0.86, `clm_0003` supported 0.86, `clm_0008` supported 0.86, `clm_0009` supported 0.86
- **能力存在：Data Validation / Profiling**（supported）：可以相信项目包含这类能力线索；是否适合你的具体任务仍要试用或安装后验证。 证据：`packages/python/goldencheck/README.md`, `packages/python/goldencheck/CLAUDE.md`, `packages/python/goldencheck/goldencheck/profilers/CLAUDE.md` Claim：`clm_0002` supported 0.86, `clm_0006` supported 0.86, `clm_0007` supported 0.86, `clm_0012` supported 0.86
- **能力存在：Data Standardization / Normalization**（supported）：可以相信项目包含这类能力线索；是否适合你的具体任务仍要试用或安装后验证。 证据：`README.md`, `examples/sql/README.md` Claim：`clm_0001` supported 0.86, `clm_0003` supported 0.86, `clm_0005` supported 0.86, `clm_0008` supported 0.86
- **能力存在：Privacy-Preserving Record Linkage (PPRL)**（supported）：可以相信项目包含这类能力线索；是否适合你的具体任务仍要试用或安装后验证。 证据：`examples/python/README.md`, `examples/airflow/README.md`, `examples/python/04_pprl_two_party.py` Claim：`clm_0004` supported 0.86, `clm_0008` supported 0.86, `clm_0010` supported 0.86, `clm_0013` supported 0.86
- **能力存在：SQL UDFs (DuckDB & PostgreSQL)**（supported）：可以相信项目包含这类能力线索；是否适合你的具体任务仍要试用或安装后验证。 证据：`examples/sql/README.md`, `examples/sql/duckdb_core_apis.sql`, `packages/rust/extensions/duckdb/README.md` Claim：`clm_0003` supported 0.86, `clm_0005` supported 0.86

### 现在还不能相信

- **角色质量和任务匹配不能直接相信。**（unverified）：角色库证明有很多角色，不证明每个角色都适合你的具体任务，也不证明角色能产生高质量结果。
- **不能把角色文案当成真实执行能力。**（unverified）：安装前只能判断角色描述和任务画像是否匹配，不能证明它能在宿主 AI 里完成任务。
- **真实输出质量不能在安装前相信。**（unverified）：Prompt Preview 只能展示引导方式，不能证明真实项目中的结果质量。
- **宿主 AI 版本兼容性不能在安装前相信。**（unverified）：Claude、Cursor、Codex、Gemini 等宿主加载规则和版本差异必须在真实环境验证。
- **不会污染现有宿主 AI 行为，不能直接相信。**（inferred）：Skill、plugin、AGENTS/CLAUDE/GEMINI 指令可能改变宿主 AI 的默认行为。 证据：`CLAUDE.md`, `packages/python/goldencheck-types/CLAUDE.md`, `packages/python/goldencheck/CLAUDE.md`, `packages/python/goldencheck/goldencheck/llm/CLAUDE.md` 等
- **可安全回滚不能默认相信。**（unverified）：除非项目明确提供卸载和恢复说明，否则必须先在隔离环境验证。
- **真实安装后是否与用户当前宿主 AI 版本兼容？**（unverified）：兼容性只能通过实际宿主环境验证。
- **项目输出质量是否满足用户具体任务？**（unverified）：安装前预览只能展示流程和边界，不能替代真实评测。

### 继续会触碰什么

- **角色选择偏差**：用户对任务应该由哪个专家角色处理的判断。 原因：选错角色会让 AI 从错误专业视角回答，浪费时间或误导决策。
- **命令执行**：包管理器、网络下载、本地插件目录、项目配置或用户主目录。 原因：运行第一条命令就可能产生环境改动；必须先判断是否值得跑。 证据：`README.md`, `packages/python/goldencheck/README.md`, `packages/python/goldencheck/examples/README.md`
- **宿主 AI 配置**：Claude/Codex/Cursor/Gemini/OpenCode 等宿主的 plugin、Skill 或规则加载配置。 原因：宿主配置会改变 AI 后续工作方式，可能和用户已有规则冲突。 证据：`CLAUDE.md`, `packages/python/goldencheck-types/CLAUDE.md`, `packages/python/goldencheck/CLAUDE.md`, `packages/python/goldencheck/goldencheck/llm/CLAUDE.md` 等
- **本地环境或项目文件**：安装结果、插件缓存、项目配置或本地依赖目录。 原因：安装前无法证明写入范围和回滚方式，需要隔离验证。 证据：`CLAUDE.md`, `README.md`, `examples/airflow/README.md`, `examples/airflow/golden_suite_identity_graph.py` 等
- **环境变量 / API Key**：项目入口文档明确出现 API key、token、secret 或账号凭证配置。 原因：如果真实安装需要凭证，应先使用测试凭证并经过权限/合规判断。 证据：`CLAUDE.md`, `packages/python/goldenmatch/scripts/bench_data_gen_server.py`, `packages/python/goldenmatch/scripts/trigger_bench_gen.py`, `packages/python/infermap/benchmark/README.md`
- **宿主 AI 上下文**：AI Context Pack、Prompt Preview、Skill 路由、风险规则和项目事实。 原因：导入上下文会影响宿主 AI 后续判断，必须避免把未验证项包装成事实。

### 最小安全下一步

- **先跑 Prompt Preview**：先用交互式试用验证任务画像和角色匹配，不要先导入整套角色库。（适用：任何项目都适用，尤其是输出质量未知时。）
- **只在隔离目录或测试账号试装**：避免安装命令污染主力宿主 AI、真实项目或用户主目录。（适用：存在命令执行、插件配置或本地写入线索时。）
- **先备份宿主 AI 配置**：Skill、plugin、规则文件可能改变 Claude/Cursor/Codex 的默认行为。（适用：存在插件 manifest、Skill 或宿主规则入口时。）
- **不要使用真实生产凭证**：环境变量/API key 一旦进入宿主或工具链，可能产生账号和合规风险。（适用：出现 API、TOKEN、KEY、SECRET 等环境线索时。）
- **安装后只验证一个最小任务**：先验证加载、兼容、输出质量和回滚，再决定是否深用。（适用：准备从试用进入真实工作流时。）

### 退出方式

- **保留安装前状态**：记录原始宿主配置和项目状态，后续才能判断是否可恢复。
- **准备移除宿主 plugin / Skill / 规则入口**：如果试装后行为异常，可以把宿主 AI 恢复到试装前状态。
- **保留原始角色选择记录**：如果输出偏题，可以回到任务画像阶段重新选择角色，而不是继续沿着错误角色推进。
- **记录安装命令和写入路径**：没有明确卸载说明时，至少要知道哪些目录或配置需要手动清理。
- **准备撤销测试 API key 或 token**：测试凭证泄露或误用时，可以快速止损。
- **如果没有回滚路径，不进入主力环境**：不可回滚是继续前阻断项，不应靠信任或运气继续。

## 哪些只能预览

- 解释项目适合谁和能做什么
- 基于项目文档演示典型对话流程
- 帮助用户判断是否值得安装或继续研究

## 哪些必须安装后验证

- 真实安装 Skill、插件或 CLI
- 执行脚本、修改本地文件或访问外部服务
- 验证真实输出质量、性能和兼容性

## 边界与风险判断卡

- **把安装前预览误认为真实运行**：用户可能高估项目已经完成的配置、权限和兼容性验证。 处理方式：明确区分 prompt_preview_can_do 与 runtime_required。 Claim：`clm_0045` inferred 0.45
- **命令执行会修改本地环境**：安装命令可能写入用户主目录、宿主插件目录或项目配置。 处理方式：先在隔离环境或测试账号中运行。 证据：`README.md`, `packages/python/goldencheck/README.md`, `packages/python/goldencheck/examples/README.md` Claim：`clm_0046` supported 0.86
- **风险**： 处理方式：
- **风险**： 处理方式：
- **风险**： 处理方式：
- **风险**： 处理方式：
- **待确认**：真实安装后是否与用户当前宿主 AI 版本兼容？。原因：兼容性只能通过实际宿主环境验证。
- **待确认**：项目输出质量是否满足用户具体任务？。原因：安装前预览只能展示流程和边界，不能替代真实评测。
- **待确认**：安装命令是否需要网络、权限或全局写入？。原因：这影响企业环境和个人环境的安装风险。

## 开工前工作上下文

### 加载顺序

- 先读取 how_to_use.host_ai_instruction，建立安装前判断资产的边界。
- 读取 claim_graph_summary，确认事实来自 Claim/Evidence Graph，而不是 Human Wiki 叙事。
- 再读取 intended_users、capabilities 和 quick_start_candidates，判断用户是否匹配。
- 需要执行具体任务时，优先查 role_skill_index，再查 evidence_index。
- 遇到真实安装、文件修改、网络访问、性能或兼容性问题时，转入 risk_card 和 boundaries.runtime_required。

### 任务路由

- **Entity Resolution / Deduplication**：先基于 role_skill_index / evidence_index 帮用户挑选可用角色、Skill 或工作流。 边界：可做安装前 Prompt 体验。 证据：`packages/python/goldenmatch/README.md`, `README.md`, `docs/adr/README.md` Claim：`clm_0001` supported 0.86, `clm_0003` supported 0.86, `clm_0008` supported 0.86, `clm_0009` supported 0.86 等
- **Data Validation / Profiling**：先基于 role_skill_index / evidence_index 帮用户挑选可用角色、Skill 或工作流。 边界：可做安装前 Prompt 体验。 证据：`packages/python/goldencheck/README.md`, `packages/python/goldencheck/CLAUDE.md`, `packages/python/goldencheck/goldencheck/profilers/CLAUDE.md` Claim：`clm_0002` supported 0.86, `clm_0006` supported 0.86, `clm_0007` supported 0.86, `clm_0012` supported 0.86 等
- **Data Standardization / Normalization**：先基于 role_skill_index / evidence_index 帮用户挑选可用角色、Skill 或工作流。 边界：可做安装前 Prompt 体验。 证据：`README.md`, `examples/sql/README.md` Claim：`clm_0001` supported 0.86, `clm_0003` supported 0.86, `clm_0005` supported 0.86, `clm_0008` supported 0.86 等
- **Privacy-Preserving Record Linkage (PPRL)**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`examples/python/README.md`, `examples/airflow/README.md`, `examples/python/04_pprl_two_party.py` Claim：`clm_0004` supported 0.86, `clm_0008` supported 0.86, `clm_0010` supported 0.86, `clm_0013` supported 0.86 等
- **SQL UDFs (DuckDB & PostgreSQL)**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`examples/sql/README.md`, `examples/sql/duckdb_core_apis.sql`, `packages/rust/extensions/duckdb/README.md` Claim：`clm_0003` supported 0.86, `clm_0005` supported 0.86
- **CLI Commands**：先基于 role_skill_index / evidence_index 帮用户挑选可用角色、Skill 或工作流。 边界：可做安装前 Prompt 体验。 证据：`packages/python/goldencheck/CLAUDE.md`, `examples/typescript/README.md` Claim：`clm_0002` supported 0.86, `clm_0006` supported 0.86, `clm_0007` supported 0.86, `clm_0011` supported 0.86 等
- **Interactive TUI**：先基于 role_skill_index / evidence_index 帮用户挑选可用角色、Skill 或工作流。 边界：可做安装前 Prompt 体验。 证据：`packages/python/goldencheck/README.md`, `packages/python/goldencheck/CLAUDE.md` Claim：`clm_0002` supported 0.86, `clm_0006` supported 0.86, `clm_0007` supported 0.86, `clm_0012` supported 0.86 等
- **Web UI Workbench**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`README.md`, `examples/python/README.md` Claim：`clm_0001` supported 0.86, `clm_0003` supported 0.86, `clm_0004` supported 0.86, `clm_0008` supported 0.86 等
- **MCP Server**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`README.md`, `CLAUDE.md`, `packages/python/goldencheck/goldencheck/mcp/CLAUDE.md`, `examples/python/06_mcp_client.py` Claim：`clm_0001` supported 0.86, `clm_0003` supported 0.86, `clm_0008` supported 0.86, `clm_0009` supported 0.86 等
- **Airflow Integration**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`examples/airflow/README.md`, `examples/airflow/golden_suite_identity_graph.py` Claim：`clm_0004` supported 0.86, `clm_0010` supported 0.86, `clm_0013` supported 0.86, `clm_0014` supported 0.86 等
- **TypeScript / Edge Runtime Support**：先基于 role_skill_index / evidence_index 帮用户挑选可用角色、Skill 或工作流。 边界：可做安装前 Prompt 体验。 证据：`examples/typescript/README.md`, `examples/typescript/02-edge-runtime.ts` Claim：`clm_0006` supported 0.86, `clm_0011` supported 0.86
- **LLM Enhancement**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`packages/python/goldencheck/goldencheck/llm/CLAUDE.md`, `packages/python/goldencheck/CLAUDE.md` Claim：`clm_0002` supported 0.86, `clm_0006` supported 0.86, `clm_0007` supported 0.86, `clm_0012` supported 0.86 等
- **Identity Graph**：先基于 role_skill_index / evidence_index 帮用户挑选可用角色、Skill 或工作流。 边界：可做安装前 Prompt 体验。 证据：`examples/python/README.md`, `examples/airflow/README.md` Claim：`clm_0004` supported 0.86, `clm_0008` supported 0.86, `clm_0010` supported 0.86, `clm_0013` supported 0.86 等
- **Review Workflow / Active Learning**：先基于 role_skill_index / evidence_index 帮用户挑选可用角色、Skill 或工作流。 边界：可做安装前 Prompt 体验。 证据：`examples/python/README.md`, `examples/airflow/README.md` Claim：`clm_0004` supported 0.86, `clm_0008` supported 0.86, `clm_0010` supported 0.86, `clm_0013` supported 0.86 等
- **Schema Mapping (InferMap)**：先基于 role_skill_index / evidence_index 帮用户挑选可用角色、Skill 或工作流。 边界：可做安装前 Prompt 体验。 证据：`examples/python/README.md`, `examples/airflow/README.md` Claim：`clm_0004` supported 0.86, `clm_0008` supported 0.86, `clm_0010` supported 0.86, `clm_0013` supported 0.86 等
- **Baseline / Drift Detection**：先基于 role_skill_index / evidence_index 帮用户挑选可用角色、Skill 或工作流。 边界：可做安装前 Prompt 体验。 证据：`packages/python/goldencheck/CLAUDE.md`, `packages/python/goldencheck/goldencheck/profilers/CLAUDE.md` Claim：`clm_0002` supported 0.86, `clm_0006` supported 0.86, `clm_0007` supported 0.86, `clm_0012` supported 0.86 等
- **Domain Packs**：先基于 role_skill_index / evidence_index 帮用户挑选可用角色、Skill 或工作流。 边界：可做安装前 Prompt 体验。 证据：`packages/python/goldencheck-types/CLAUDE.md`, `packages/python/goldencheck/README.md`, `packages/python/goldencheck/goldencheck/mcp/CLAUDE.md` Claim：`clm_0002` supported 0.86, `clm_0007` supported 0.86, `clm_0009` supported 0.86, `clm_0017` supported 0.86
- **GitHub Actions**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`packages/actions/goldencheck/README.md`, `packages/actions/goldenflow/README.md`, `packages/actions/goldenmatch/README.md` Claim：`clm_0018` supported 0.86
- **Auto-Configuration (Autoconfig)**：先基于 role_skill_index / evidence_index 帮用户挑选可用角色、Skill 或工作流。 边界：可做安装前 Prompt 体验。 证据：`docs/adr/README.md`, `docs/adr/2026-05-21-v1-13-autoconfig-roadmap.md` Claim：`clm_0001` supported 0.86, `clm_0019` supported 0.86

### 上下文规模

- 文件总数：2823
- 重要文件覆盖：40/2823
- 证据索引条目：80
- 角色 / Skill 条目：67

### 证据不足时的处理

- **missing_evidence**：说明证据不足，要求用户提供目标文件、README 段落或安装后验证记录；不要补全事实。
- **out_of_scope_request**：说明该任务超出当前 AI Context Pack 证据范围，并建议用户先查看 Human Manual 或真实安装后验证。
- **runtime_request**：给出安装前检查清单和命令来源，但不要替用户执行命令或声称已执行。
- **source_conflict**：同时展示冲突来源，标记为待核实，不要强行选择一个版本。

## Prompt Recipes

### 适配判断

- 目标：判断这个项目是否适合用户当前任务。
- 预期输出：适配结论、关键理由、证据引用、安装前可预览内容、必须安装后验证内容、下一步建议。

```text
请基于 goldenmatch-monorepo 的 AI Context Pack，先问我 3 个必要问题，然后判断它是否适合我的任务。回答必须包含：适合谁、能做什么、不能做什么、是否值得安装、证据来自哪里。所有项目事实必须引用 evidence_refs、source_paths 或 claim_id。
```

### 安装前体验

- 目标：让用户在安装前感受核心工作流，同时避免把预览包装成真实能力或营销承诺。
- 预期输出：一段带边界标签的体验剧本、安装后验证清单和谨慎建议；不含真实运行承诺或强营销表述。

```text
请把 goldenmatch-monorepo 当作安装前体验资产，而不是已安装工具或真实运行环境。

请严格输出四段：
1. 先问我 3 个必要问题。
2. 给出一段“体验剧本”：用 [安装前可预览]、[必须安装后验证]、[证据不足] 三种标签展示它可能如何引导工作流。
3. 给出安装后验证清单：列出哪些能力只有真实安装、真实宿主加载、真实项目运行后才能确认。
4. 给出谨慎建议：只能说“值得继续研究/试装”“先补充信息后再判断”或“不建议继续”，不得替项目背书。

硬性边界：
- 不要声称已经安装、运行、执行测试、修改文件或产生真实结果。
- 不要写“自动适配”“确保通过”“完美适配”“强烈建议安装”等承诺性表达。
- 如果描述安装后的工作方式，必须使用“如果安装成功且宿主正确加载 Skill，它可能会……”这种条件句。
- 体验剧本只能写成“示例台词/假设流程”：使用“可能会询问/可能会建议/可能会展示”，不要写“已写入、已生成、已通过、正在运行、正在生成”。
- Prompt Preview 不负责给安装命令；如用户准备试装，只能提示先阅读 Quick Start 和 Risk Card，并在隔离环境验证。
- 所有项目事实必须来自 supported claim、evidence_refs 或 source_paths；inferred/unverified 只能作风险或待确认项。

```

### 角色 / Skill 选择

- 目标：从项目里的角色或 Skill 中挑选最匹配的资产。
- 预期输出：候选角色或 Skill 列表，每项包含适用场景、证据路径、风险边界和是否需要安装后验证。

```text
请读取 role_skill_index，根据我的目标任务推荐 3-5 个最相关的角色或 Skill。每个推荐都要说明适用场景、可能输出、风险边界和 evidence_refs。
```

### 风险预检

- 目标：安装或引入前识别环境、权限、规则冲突和质量风险。
- 预期输出：环境、权限、依赖、许可、宿主冲突、质量风险和未知项的检查清单。

```text
请基于 risk_card、boundaries 和 quick_start_candidates，给我一份安装前风险预检清单。不要替我执行命令，只说明我应该检查什么、为什么检查、失败会有什么影响。
```

### 宿主 AI 开工指令

- 目标：把项目上下文转成一次对话开始前的宿主 AI 指令。
- 预期输出：一段边界明确、证据引用明确、适合复制给宿主 AI 的开工前指令。

```text
请基于 goldenmatch-monorepo 的 AI Context Pack，生成一段我可以粘贴给宿主 AI 的开工前指令。这段指令必须遵守 not_runtime=true，不能声称项目已经安装、运行或产生真实结果。
```


## 角色 / Skill 索引

- 共索引 67 个角色 / Skill / 项目文档条目。

- **Architecture Decision Records ADRs**（project_doc）：Captures load-bearing architectural decisions in goldenmatch. Each ADR is a small markdown file in this directory, numbered sequentially. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/adr/README.md`
- **superpowers workflow notes**（project_doc）：docs/superpowers/specs/ and docs/superpowers/plans/ are matched by .gitignore . Use git add -f to commit them. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/superpowers/CLAUDE.md`
- **Golden Suite monorepo**（project_doc）：Polyglot monorepo: packages/{python,rust,typescript,dbt,actions} . Per-package CLAUDE.md files own package-specific context. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`CLAUDE.md`
- **🟡 Golden Suite**（project_doc）：A polyglot data-quality and entity-resolution toolkit. Polished, opinionated, AI-native. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`README.md`
- **Examples**（project_doc）：Runnable demos for the Golden Suite, organized by host. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`examples/README.md`
- **Airflow DAGs for the Golden Suite**（project_doc）：Drop-in DAG examples that wire the Golden Suite into a real Airflow deployment. Copy the file you want into your Airflow dags/ folder, adjust the knobs at the top, and ship. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`examples/airflow/README.md`
- **Python usage examples**（project_doc）：Cross-suite, runnable scripts. Each is standalone — pick the one closest to your scenario, adapt to your data shape, ship. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`examples/python/README.md`
- **SQL usage examples**（project_doc）：Run GoldenMatch entity resolution directly from SQL. Both backends expose the same surface: 13 core-API functions goldenmatch plus 8 GoldenFlow transforms goldenflow , on top of the pre-existing dedupe / match / score / identity functions. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`examples/sql/README.md`
- **TypeScript usage examples**（project_doc）：File What Imports --- --- --- 01-quickstart.ts 30-second dedupe of a record array goldenmatch 02-edge-runtime.ts Vercel Edge / Cloudflare Workers route. Uses goldenmatch/core no node: imports . goldenmatch/core 03-mcp-client.ts Connect to the goldensuite-mcp container from a TS MCP client. @modelcontextprotocol/sdk 04-goldenpipe-orchestration.ts Chain check - flow - dedupe through the goldenpipe orchestrator via run… 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`examples/typescript/README.md`
- **GoldenCheck Action**（project_doc）：GitHub Action for GoldenCheck https://github.com/benseverndev-oss/goldencheck — data validation that discovers rules from your data. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/actions/goldencheck/README.md`
- **GoldenFlow Action**（project_doc）：Run GoldenFlow https://github.com/benseverndev-oss/goldenmatch data transformations on your data files in CI, report what changed, and post a PR comment. Companion to the GoldenCheck action. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/actions/goldenflow/README.md`
- **GoldenMatch Action**（project_doc）：Deduplicate data files in CI with GoldenMatch https://github.com/benseverndev-oss/goldenmatch , report cluster/duplicate counts, gate on a duplicate threshold, and post a PR comment. Companion to the GoldenCheck and GoldenFlow actions. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/actions/goldenmatch/README.md`
- **Python packages — Claude notes**（project_doc）：- uv workspace rooted at /pyproject.toml . Members listed in tool.uv.workspace and tool.uv.sources . Both must be updated when adding a package. - Cross-package deps use goldencheck-types -style names no @workspace/ prefix . All resolved via tool.uv.sources = { workspace = true } at the root. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/CLAUDE.md`
- **goldencheck-types**（project_doc）：Shared canonical field-type registry for the Golden Suite. Producer: packages/python/goldencheck/ . Consumers: packages/python/goldenpipe/ stage I/O , packages/python/infermap/ target schema , packages/typescript/goldencheck-types/ cross-language mirror over JSON wire . 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldencheck-types/CLAUDE.md`
- **GoldenCheck**（project_doc）：Data validation that discovers rules from your data. DQBench Score: 88.40. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldencheck/AGENTS.md`
- **GoldenCheck**（project_doc）：Data validation that discovers rules from your data. DQBench Score: 88.40. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldencheck/CLAUDE.md`
- **GoldenCheck**（project_doc）：Data validation that discovers rules from your data so you don't have to write them. Built by Ben Severn https://bensevern.dev . 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldencheck/README.md`
- **GoldenCheck Examples**（project_doc）：Script What It Does Prerequisites -------- ------------- -------------- scan and profile.py Scan a CSV for issues, get health score goldencheck scan basic.py Scan a CSV file and print all findings goldencheck validate rules.py Validate data against a goldencheck.yml config goldencheck domain packs.py List and use industry-specific domain packs goldencheck domain pack.py Scan with the healthcare domain pack for clini… 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldencheck/examples/README.md`
- **Engine**（project_doc）：scan file scan file with llm --- --- --- Returns findings, profile or findings, profile, sample findings, profile return sample=True Returns 3-tuple Called internally Confidence downgrade Caller must call apply confidence downgrade Done inside LLM path Suppression Yes always Yes inside scan file 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldencheck/goldencheck/engine/CLAUDE.md`
- **LLM Module**（project_doc）：Call scan file with llm path, provider — never call stage 2 steps manually. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldencheck/goldencheck/llm/CLAUDE.md`
- **MCP Server**（project_doc）：Tool Description ------ ------------- scan Scan a file for data quality issues optional LLM boost + domain validate Check against pinned rules in goldencheck.yml profile Column-level statistics and health score health score Quick A-F grade get column detail Deep-dive into a specific column list checks Available profiler checks list domains Available domain packs bundled + community get domain info Types in a specifi… 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldencheck/goldencheck/mcp/CLAUDE.md`
- **Profilers**（project_doc）：- df is always the sampled DataFrame up to 100k rows , not the full file - context is a shared mutable dict passed to every profiler per column — use it to share intermediate results e.g., TypeInferenceProfiler writes inferred type; others read it - Return if nothing to report — never raise, log instead 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldencheck/goldencheck/profilers/CLAUDE.md`
- **Semantic**（project_doc）：Classifies each column's semantic type email, phone, date, etc. and suppresses irrelevant findings based on that type. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldencheck/goldencheck/semantic/CLAUDE.md`
- **TUI**（project_doc）：Built with Textual https://textual.textualize.io/ . Entry point: GoldenCheckApp in app.py . 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldencheck/goldencheck/tui/CLAUDE.md`
- **GoldenFlow**（project_doc）：Data transformation toolkit -- standardize, reshape, and normalize messy data. DQBench Transform Score: 100/100. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldenflow/AGENTS.md`
- **GoldenFlow**（project_doc）：Data transformation toolkit -- standardize, reshape, and normalize messy data. DQBench Transform Score: 100/100. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldenflow/CLAUDE.md`
- **GoldenFlow**（project_doc）：Data transformation toolkit — standardize, reshape, and normalize messy data before it hits your pipeline. Built by Ben Severn https://bensevern.dev . 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldenflow/README.md`
- **GoldenFlow Examples**（project_doc）：Script Description Prerequisites -------- ------------- --------------- zero config.py Zero-config transform on messy data -- auto-detect and fix goldenflow configured transform.py Explicit transforms via GoldenFlowConfig goldenflow schema mapping.py Map columns between source and target schemas goldenflow benchmark.py Run DQBench Transform benchmark score: 100.00 goldenflow , dqbench transform basic.py Zero-config… 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldenflow/examples/README.md`
- **GoldenMatch**（project_doc）：Related Projects - SQL Extensions repo: D:\show case\goldenmatch-extensions -- Postgres extension + DuckDB UDFs. Has its own CLAUDE.md. - PyPI: goldenmatch Python toolkit , goldenmatch-duckdb DuckDB UDFs - GitHub: benseverndev-oss/goldenmatch , benseverndev-oss/goldenmatch-extensions 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldenmatch/AGENTS.md`
- **GoldenMatch**（project_doc）：Related Projects - SQL Extensions repo: D:\show case\goldenmatch-extensions -- Postgres extension + DuckDB UDFs. Has its own CLAUDE.md. - PyPI: goldenmatch Python toolkit , goldenmatch-duckdb DuckDB UDFs - npm: goldenmatch TypeScript port at packages/goldenmatch-js/ - GitHub: benseverndev-oss/goldenmatch , benseverndev-oss/goldenmatch-extensions 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldenmatch/CLAUDE.md`
- **🟡 GoldenMatch**（project_doc）：Find duplicate records in 30 seconds. No rules to write, no models to train. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldenmatch/README.md`
- **dbt-goldensuite**（project_doc）：dbt integration for GoldenMatch https://github.com/benseverndev-oss/goldenmatch entity resolution. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldenmatch/dbt-goldensuite/README.md`
- **GoldenMatch Examples**（project_doc）：Self-contained, runnable scripts demonstrating every major feature. Each generates its own sample data and runs in under 10 seconds. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldenmatch/examples/README.md`
- **GoldenMatch DB migrations**（project_doc）：SQL migration scripts that mirror the schemas embedded in the Python IdentityStore / MemoryStore classes. Apply manually when you want the shared Postgres state set up by a DBA rather than by the Python process. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldenmatch/goldenmatch/db/migrations/README.md`
- **React + TypeScript + Vite**（project_doc）：This template provides a minimal setup to get React working in Vite with HMR and some ESLint rules. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldenmatch/web/frontend/README.md`
- **GoldenPipe**（project_doc）：Golden Suite orchestrator -- chains GoldenCheck, GoldenFlow, GoldenMatch. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldenpipe/AGENTS.md`
- **GoldenPipe**（project_doc）：Golden Suite orchestrator -- chains GoldenCheck, GoldenFlow, GoldenMatch. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldenpipe/CLAUDE.md`
- **GoldenPipe**（project_doc）：Golden Suite orchestrator -- Check quality, fix issues, deduplicate records. One command. Built by Ben Severn https://bensevern.dev . 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldenpipe/README.md`
- **GoldenPipe Examples**（project_doc）：Script What It Does Prerequisites -------- ------------- -------------- full suite demo.py Each Golden Suite tool individually, then the full pipeline goldenpipe golden-suite benchmark suite.py DQBench scores for all 4 tools with visual scorecard goldenpipe golden-suite dqbench basic pipeline.py Run a full pipeline on a CSV file goldenpipe golden-suite selective stages.py Run only check + flow, skip match goldenpipe… 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldenpipe/examples/README.md`
- **goldenpipe stages**（project_doc）：python from goldenpipe.models.context import PipeContext, StageResult, StageStatus from goldenpipe.models.stage import stage 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldenpipe/goldenpipe/stages/CLAUDE.md`
- **goldensuite-mcp**（project_doc）：One MCP server exposing every Golden Suite tool — goldenmatch , goldencheck , goldenflow , goldenpipe , infermap — under a single endpoint. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldensuite-mcp/README.md`
- **infermap**（project_doc）：Environment - Windows 11, bash shell Git Bash — use Unix paths in scripts - Python 3.12 at C:\Users\bsevern\AppData\Local\Programs\Python\Python312\python.exe - Project lives in the goldenmatch monorepo at packages/python/infermap/ . Pre-fold standalone path was D:\show case\infermap — archive/goldenmatch-pre-fold/ retains that history. - Two GitHub accounts: benzsevern owner and benzsevern-mjh work - Always gh auth… 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/infermap/CLAUDE.md`
- **Table of contents**（project_doc）：Inference-driven schema mapping engine. Map messy source columns to a known target schema — accurately, explainably, with zero config. Built by Ben Severn . 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/infermap/README.md`
- **infermap benchmark**（project_doc）：A 200-case accuracy benchmark for the infermap schema mapping engine. Two parallel runners Python + TypeScript consume a shared corpus, emit identical-schema report.json files, and feed a Python aggregator that posts a sticky PR comment on every qualifying pull request. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/infermap/benchmark/README.md`
- **infermap-bench**（project_doc）：Installable accuracy benchmark runner for the infermap ../../../infermap/ package. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/infermap/benchmark/runners/python/README.md`
- **infermap Examples**（project_doc）：Hands-on examples to try infermap. Each example is self-contained — just run it. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/infermap/examples/README.md`
- **Example YAML Configs**（project_doc）：Pre-built infermap configurations for common industry verticals. Each config extends the built-in alias registry with domain-specific field name synonyms and tunes scorer weights for the domain. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/infermap/examples/configs/README.md`
- **infermap — TypeScript Examples**（project_doc）：Runnable TypeScript examples for the infermap npm package. Each example is self-contained and progressively demonstrates more advanced usage. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/infermap/examples/typescript/README.md`
- **GoldenMatch Extensions**（project_doc）：Native SQL extensions for GoldenMatch https://github.com/benseverndev-oss/goldenmatch D:\show case\goldenmatch . 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/rust/extensions/AGENTS.md`
- **GoldenMatch Extensions**（project_doc）：Native SQL extensions for GoldenMatch https://github.com/benseverndev-oss/goldenmatch D:\show case\goldenmatch . 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/rust/extensions/CLAUDE.md`
- **goldenmatch-extensions**（project_doc）：Native SQL extensions for GoldenMatch https://github.com/benseverndev-oss/goldenmatch -- run entity resolution directly from PostgreSQL and DuckDB. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/rust/extensions/README.md`
- **goldenmatch-duckdb**（project_doc）：GoldenMatch entity resolution functions for DuckDB. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/rust/extensions/duckdb/README.md`
- **goldenmatch-native**（project_doc）：Optional native Rust/PyO3 acceleration kernels for goldenmatch https://github.com/benseverndev-oss/goldenmatch . 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/rust/extensions/native/README.md`
- **goldenmatch pg**（project_doc）：GoldenMatch entity resolution functions for PostgreSQL, built with pgrx https://github.com/pgcentralfoundation/pgrx . The extension embeds a CPython interpreter and calls the goldenmatch https://pypi.org/project/goldenmatch/ Python package, so install that into the same Python the server uses. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/rust/extensions/postgres/README.md`
- **TypeScript packages — Claude notes**（project_doc）：Each package has its own package.json and installs independently. This is intentional — Windows symlinks fail with EISDIR on file:../sibling paths. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/typescript/CLAUDE.md`
- **GoldenCheck Community Types**（project_doc）：Community-contributed semantic type definitions for GoldenCheck https://github.com/benseverndev-oss/goldencheck . 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/typescript/goldencheck-types/README.md`
- **goldenmatch TypeScript**（project_doc）：npm package goldenmatch . Parity port of the Python sibling at packages/python/goldenmatch/ . Currently at v0.11.0 core-algorithm parity catch-up + Phase-5 golden-strategy plugin port . Python sibling is at v1.16; v1.13/v1.14/v1.16 are explicitly not-ported see CHANGELOG.md for the per-version rationale . 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/typescript/goldenmatch/CLAUDE.md`
- **GoldenMatch TypeScript**（project_doc）：Entity resolution toolkit for Node.js and edge runtimes. Deduplicate, match, and create golden records — in TypeScript. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/typescript/goldenmatch/README.md`
- **GoldenMatch TypeScript Examples**（project_doc）：Each example is a standalone .ts file runnable with: 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/typescript/goldenmatch/examples/README.md`
- **goldenpipe TypeScript**（project_doc）：npm package goldenpipe . Port of the Python sibling at packages/python/goldenpipe/ . Suite orchestrator: chains goldencheck - goldenflow - goldenmatch the TS siblings under packages/typescript/ . Currently v0.1.0 initial port; v1 scope = the core check- flow- dedupe chain . 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/typescript/goldenpipe/CLAUDE.md`
- **goldenpipe**（project_doc）：Golden Suite orchestrator for TypeScript — chains GoldenCheck → GoldenFlow → GoldenMatch into one adaptive, pluggable pipeline. TypeScript port of the goldenpipe https://github.com/benseverndev-oss/goldenmatch/tree/main/packages/python/goldenpipe Python library. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/typescript/goldenpipe/README.md`
- **infermap**（project_doc）：Map messy source columns to a known target schema — accurately, explainably, with zero config. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/typescript/infermap/README.md`
- **Contributing to GoldenCheck**（project_doc）：Thanks for your interest in improving GoldenCheck! 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldencheck/CONTRIBUTING.md`
- **Contributing to GoldenFlow**（project_doc）：Thanks for your interest in improving GoldenFlow! 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldenflow/CONTRIBUTING.md`
- **Contributing to GoldenMatch**（project_doc）：Thanks for your interest in contributing! Here's how to get started. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldenmatch/CONTRIBUTING.md`
- **Contributing to GoldenPipe**（project_doc）：Thanks for your interest in contributing! Here's how to get started. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/goldenpipe/CONTRIBUTING.md`
- **Contributing to infermap**（project_doc）：Thanks for your interest in improving infermap! 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`packages/python/infermap/CONTRIBUTING.md`

## 证据索引

- 共索引 80 条证据。

- **Architecture Decision Records ADRs**（documentation）：Captures load-bearing architectural decisions in goldenmatch. Each ADR is a small markdown file in this directory, numbered sequentially. 证据：`docs/adr/README.md`
- **superpowers workflow notes**（documentation）：docs/superpowers/specs/ and docs/superpowers/plans/ are matched by .gitignore . Use git add -f to commit them. 证据：`docs/superpowers/CLAUDE.md`
- **Golden Suite monorepo**（documentation）：Polyglot monorepo: packages/{python,rust,typescript,dbt,actions} . Per-package CLAUDE.md files own package-specific context. 证据：`CLAUDE.md`
- **🟡 Golden Suite**（documentation）：A polyglot data-quality and entity-resolution toolkit. Polished, opinionated, AI-native. 证据：`README.md`
- **Examples**（documentation）：Runnable demos for the Golden Suite, organized by host. 证据：`examples/README.md`
- **Airflow DAGs for the Golden Suite**（documentation）：Drop-in DAG examples that wire the Golden Suite into a real Airflow deployment. Copy the file you want into your Airflow dags/ folder, adjust the knobs at the top, and ship. 证据：`examples/airflow/README.md`
- **Python usage examples**（documentation）：Cross-suite, runnable scripts. Each is standalone — pick the one closest to your scenario, adapt to your data shape, ship. 证据：`examples/python/README.md`
- **SQL usage examples**（documentation）：Run GoldenMatch entity resolution directly from SQL. Both backends expose the same surface: 13 core-API functions goldenmatch plus 8 GoldenFlow transforms goldenflow , on top of the pre-existing dedupe / match / score / identity functions. 证据：`examples/sql/README.md`
- **TypeScript usage examples**（documentation）：File What Imports --- --- --- 01-quickstart.ts 30-second dedupe of a record array goldenmatch 02-edge-runtime.ts Vercel Edge / Cloudflare Workers route. Uses goldenmatch/core no node: imports . goldenmatch/core 03-mcp-client.ts Connect to the goldensuite-mcp container from a TS MCP client. @modelcontextprotocol/sdk 04-goldenpipe-orchestration.ts Chain check - flow - dedupe through the goldenpipe orchestrator via runDf zero-config + custom config . goldenpipe 证据：`examples/typescript/README.md`
- **GoldenCheck Action**（documentation）：GitHub Action for GoldenCheck https://github.com/benseverndev-oss/goldencheck — data validation that discovers rules from your data. 证据：`packages/actions/goldencheck/README.md`
- **GoldenFlow Action**（documentation）：Run GoldenFlow https://github.com/benseverndev-oss/goldenmatch data transformations on your data files in CI, report what changed, and post a PR comment. Companion to the GoldenCheck action. 证据：`packages/actions/goldenflow/README.md`
- **GoldenMatch Action**（documentation）：Deduplicate data files in CI with GoldenMatch https://github.com/benseverndev-oss/goldenmatch , report cluster/duplicate counts, gate on a duplicate threshold, and post a PR comment. Companion to the GoldenCheck and GoldenFlow actions. 证据：`packages/actions/goldenmatch/README.md`
- **Python packages — Claude notes**（documentation）：- uv workspace rooted at /pyproject.toml . Members listed in tool.uv.workspace and tool.uv.sources . Both must be updated when adding a package. - Cross-package deps use goldencheck-types -style names no @workspace/ prefix . All resolved via tool.uv.sources = { workspace = true } at the root. 证据：`packages/python/CLAUDE.md`
- **goldencheck-types**（documentation）：Shared canonical field-type registry for the Golden Suite. Producer: packages/python/goldencheck/ . Consumers: packages/python/goldenpipe/ stage I/O , packages/python/infermap/ target schema , packages/typescript/goldencheck-types/ cross-language mirror over JSON wire . 证据：`packages/python/goldencheck-types/CLAUDE.md`
- **GoldenCheck**（documentation）：Data validation that discovers rules from your data. DQBench Score: 88.40. 证据：`packages/python/goldencheck/AGENTS.md`
- **GoldenCheck**（documentation）：Data validation that discovers rules from your data. DQBench Score: 88.40. 证据：`packages/python/goldencheck/CLAUDE.md`
- **GoldenCheck**（documentation）：Data validation that discovers rules from your data so you don't have to write them. Built by Ben Severn https://bensevern.dev . 证据：`packages/python/goldencheck/README.md`
- **GoldenCheck Examples**（documentation）：Script What It Does Prerequisites -------- ------------- -------------- scan and profile.py Scan a CSV for issues, get health score goldencheck scan basic.py Scan a CSV file and print all findings goldencheck validate rules.py Validate data against a goldencheck.yml config goldencheck domain packs.py List and use industry-specific domain packs goldencheck domain pack.py Scan with the healthcare domain pack for clinical data types goldencheck benchmark.py Run DQBench Detect benchmark goldencheck dqbench 证据：`packages/python/goldencheck/examples/README.md`
- **Engine**（documentation）：scan file scan file with llm --- --- --- Returns findings, profile or findings, profile, sample findings, profile return sample=True Returns 3-tuple Called internally Confidence downgrade Caller must call apply confidence downgrade Done inside LLM path Suppression Yes always Yes inside scan file 证据：`packages/python/goldencheck/goldencheck/engine/CLAUDE.md`
- **LLM Module**（documentation）：Call scan file with llm path, provider — never call stage 2 steps manually. 证据：`packages/python/goldencheck/goldencheck/llm/CLAUDE.md`
- **MCP Server**（documentation）：Tool Description ------ ------------- scan Scan a file for data quality issues optional LLM boost + domain validate Check against pinned rules in goldencheck.yml profile Column-level statistics and health score health score Quick A-F grade get column detail Deep-dive into a specific column list checks Available profiler checks list domains Available domain packs bundled + community get domain info Types in a specific domain pack install domain Download community domain pack to goldencheck domain.yaml 证据：`packages/python/goldencheck/goldencheck/mcp/CLAUDE.md`
- **Profilers**（documentation）：- df is always the sampled DataFrame up to 100k rows , not the full file - context is a shared mutable dict passed to every profiler per column — use it to share intermediate results e.g., TypeInferenceProfiler writes inferred type; others read it - Return if nothing to report — never raise, log instead 证据：`packages/python/goldencheck/goldencheck/profilers/CLAUDE.md`
- **Semantic**（documentation）：Classifies each column's semantic type email, phone, date, etc. and suppresses irrelevant findings based on that type. 证据：`packages/python/goldencheck/goldencheck/semantic/CLAUDE.md`
- **TUI**（documentation）：Built with Textual https://textual.textualize.io/ . Entry point: GoldenCheckApp in app.py . 证据：`packages/python/goldencheck/goldencheck/tui/CLAUDE.md`
- **GoldenFlow**（documentation）：Data transformation toolkit -- standardize, reshape, and normalize messy data. DQBench Transform Score: 100/100. 证据：`packages/python/goldenflow/AGENTS.md`
- **GoldenFlow**（documentation）：Data transformation toolkit -- standardize, reshape, and normalize messy data. DQBench Transform Score: 100/100. 证据：`packages/python/goldenflow/CLAUDE.md`
- **GoldenFlow**（documentation）：Data transformation toolkit — standardize, reshape, and normalize messy data before it hits your pipeline. Built by Ben Severn https://bensevern.dev . 证据：`packages/python/goldenflow/README.md`
- **GoldenFlow Examples**（documentation）：Script Description Prerequisites -------- ------------- --------------- zero config.py Zero-config transform on messy data -- auto-detect and fix goldenflow configured transform.py Explicit transforms via GoldenFlowConfig goldenflow schema mapping.py Map columns between source and target schemas goldenflow benchmark.py Run DQBench Transform benchmark score: 100.00 goldenflow , dqbench transform basic.py Zero-config file transform with manifest output goldenflow config based.py Transform with a YAML config for explicit control goldenflow domain pack.py Healthcare domain transforms for clinical data goldenflow 证据：`packages/python/goldenflow/examples/README.md`
- **GoldenMatch**（documentation）：Related Projects - SQL Extensions repo: D:\show case\goldenmatch-extensions -- Postgres extension + DuckDB UDFs. Has its own CLAUDE.md. - PyPI: goldenmatch Python toolkit , goldenmatch-duckdb DuckDB UDFs - GitHub: benseverndev-oss/goldenmatch , benseverndev-oss/goldenmatch-extensions 证据：`packages/python/goldenmatch/AGENTS.md`
- **GoldenMatch**（documentation）：Related Projects - SQL Extensions repo: D:\show case\goldenmatch-extensions -- Postgres extension + DuckDB UDFs. Has its own CLAUDE.md. - PyPI: goldenmatch Python toolkit , goldenmatch-duckdb DuckDB UDFs - npm: goldenmatch TypeScript port at packages/goldenmatch-js/ - GitHub: benseverndev-oss/goldenmatch , benseverndev-oss/goldenmatch-extensions 证据：`packages/python/goldenmatch/CLAUDE.md`
- **🟡 GoldenMatch**（documentation）：Find duplicate records in 30 seconds. No rules to write, no models to train. 证据：`packages/python/goldenmatch/README.md`
- **dbt-goldensuite**（documentation）：dbt integration for GoldenMatch https://github.com/benseverndev-oss/goldenmatch entity resolution. 证据：`packages/python/goldenmatch/dbt-goldensuite/README.md`
- **GoldenMatch Examples**（documentation）：Self-contained, runnable scripts demonstrating every major feature. Each generates its own sample data and runs in under 10 seconds. 证据：`packages/python/goldenmatch/examples/README.md`
- **GoldenMatch DB migrations**（documentation）：SQL migration scripts that mirror the schemas embedded in the Python IdentityStore / MemoryStore classes. Apply manually when you want the shared Postgres state set up by a DBA rather than by the Python process. 证据：`packages/python/goldenmatch/goldenmatch/db/migrations/README.md`
- **React + TypeScript + Vite**（documentation）：This template provides a minimal setup to get React working in Vite with HMR and some ESLint rules. 证据：`packages/python/goldenmatch/web/frontend/README.md`
- **GoldenPipe**（documentation）：Golden Suite orchestrator -- chains GoldenCheck, GoldenFlow, GoldenMatch. 证据：`packages/python/goldenpipe/AGENTS.md`
- **GoldenPipe**（documentation）：Golden Suite orchestrator -- chains GoldenCheck, GoldenFlow, GoldenMatch. 证据：`packages/python/goldenpipe/CLAUDE.md`
- **GoldenPipe**（documentation）：Golden Suite orchestrator -- Check quality, fix issues, deduplicate records. One command. Built by Ben Severn https://bensevern.dev . 证据：`packages/python/goldenpipe/README.md`
- **GoldenPipe Examples**（documentation）：Script What It Does Prerequisites -------- ------------- -------------- full suite demo.py Each Golden Suite tool individually, then the full pipeline goldenpipe golden-suite benchmark suite.py DQBench scores for all 4 tools with visual scorecard goldenpipe golden-suite dqbench basic pipeline.py Run a full pipeline on a CSV file goldenpipe golden-suite selective stages.py Run only check + flow, skip match goldenpipe golden-suite custom pipeline.py Build a custom pipeline with specific stages goldenpipe golden-suite 证据：`packages/python/goldenpipe/examples/README.md`
- **goldenpipe stages**（documentation）：python from goldenpipe.models.context import PipeContext, StageResult, StageStatus from goldenpipe.models.stage import stage 证据：`packages/python/goldenpipe/goldenpipe/stages/CLAUDE.md`
- **goldensuite-mcp**（documentation）：One MCP server exposing every Golden Suite tool — goldenmatch , goldencheck , goldenflow , goldenpipe , infermap — under a single endpoint. 证据：`packages/python/goldensuite-mcp/README.md`
- **infermap**（documentation）：Environment - Windows 11, bash shell Git Bash — use Unix paths in scripts - Python 3.12 at C:\Users\bsevern\AppData\Local\Programs\Python\Python312\python.exe - Project lives in the goldenmatch monorepo at packages/python/infermap/ . Pre-fold standalone path was D:\show case\infermap — archive/goldenmatch-pre-fold/ retains that history. - Two GitHub accounts: benzsevern owner and benzsevern-mjh work - Always gh auth switch --user benzsevern before push, switch back after - PyPI: infermap v0.1.0 published trusted publishing configured 证据：`packages/python/infermap/CLAUDE.md`
- **Table of contents**（documentation）：Inference-driven schema mapping engine. Map messy source columns to a known target schema — accurately, explainably, with zero config. Built by Ben Severn . 证据：`packages/python/infermap/README.md`
- **infermap benchmark**（documentation）：A 200-case accuracy benchmark for the infermap schema mapping engine. Two parallel runners Python + TypeScript consume a shared corpus, emit identical-schema report.json files, and feed a Python aggregator that posts a sticky PR comment on every qualifying pull request. 证据：`packages/python/infermap/benchmark/README.md`
- **infermap-bench**（documentation）：Installable accuracy benchmark runner for the infermap ../../../infermap/ package. 证据：`packages/python/infermap/benchmark/runners/python/README.md`
- **infermap Examples**（documentation）：Hands-on examples to try infermap. Each example is self-contained — just run it. 证据：`packages/python/infermap/examples/README.md`
- **Example YAML Configs**（documentation）：Pre-built infermap configurations for common industry verticals. Each config extends the built-in alias registry with domain-specific field name synonyms and tunes scorer weights for the domain. 证据：`packages/python/infermap/examples/configs/README.md`
- **infermap — TypeScript Examples**（documentation）：Runnable TypeScript examples for the infermap npm package. Each example is self-contained and progressively demonstrates more advanced usage. 证据：`packages/python/infermap/examples/typescript/README.md`
- **GoldenMatch Extensions**（documentation）：Native SQL extensions for GoldenMatch https://github.com/benseverndev-oss/goldenmatch D:\show case\goldenmatch . 证据：`packages/rust/extensions/AGENTS.md`
- **GoldenMatch Extensions**（documentation）：Native SQL extensions for GoldenMatch https://github.com/benseverndev-oss/goldenmatch D:\show case\goldenmatch . 证据：`packages/rust/extensions/CLAUDE.md`
- **goldenmatch-extensions**（documentation）：Native SQL extensions for GoldenMatch https://github.com/benseverndev-oss/goldenmatch -- run entity resolution directly from PostgreSQL and DuckDB. 证据：`packages/rust/extensions/README.md`
- **goldenmatch-duckdb**（documentation）：GoldenMatch entity resolution functions for DuckDB. 证据：`packages/rust/extensions/duckdb/README.md`
- **goldenmatch-native**（documentation）：Optional native Rust/PyO3 acceleration kernels for goldenmatch https://github.com/benseverndev-oss/goldenmatch . 证据：`packages/rust/extensions/native/README.md`
- **goldenmatch pg**（documentation）：GoldenMatch entity resolution functions for PostgreSQL, built with pgrx https://github.com/pgcentralfoundation/pgrx . The extension embeds a CPython interpreter and calls the goldenmatch https://pypi.org/project/goldenmatch/ Python package, so install that into the same Python the server uses. 证据：`packages/rust/extensions/postgres/README.md`
- **TypeScript packages — Claude notes**（documentation）：Each package has its own package.json and installs independently. This is intentional — Windows symlinks fail with EISDIR on file:../sibling paths. 证据：`packages/typescript/CLAUDE.md`
- **GoldenCheck Community Types**（documentation）：Community-contributed semantic type definitions for GoldenCheck https://github.com/benseverndev-oss/goldencheck . 证据：`packages/typescript/goldencheck-types/README.md`
- **goldenmatch TypeScript**（documentation）：npm package goldenmatch . Parity port of the Python sibling at packages/python/goldenmatch/ . Currently at v0.11.0 core-algorithm parity catch-up + Phase-5 golden-strategy plugin port . Python sibling is at v1.16; v1.13/v1.14/v1.16 are explicitly not-ported see CHANGELOG.md for the per-version rationale . 证据：`packages/typescript/goldenmatch/CLAUDE.md`
- **GoldenMatch TypeScript**（documentation）：Entity resolution toolkit for Node.js and edge runtimes. Deduplicate, match, and create golden records — in TypeScript. 证据：`packages/typescript/goldenmatch/README.md`
- **GoldenMatch TypeScript Examples**（documentation）：Each example is a standalone .ts file runnable with: 证据：`packages/typescript/goldenmatch/examples/README.md`
- **goldenpipe TypeScript**（documentation）：npm package goldenpipe . Port of the Python sibling at packages/python/goldenpipe/ . Suite orchestrator: chains goldencheck - goldenflow - goldenmatch the TS siblings under packages/typescript/ . Currently v0.1.0 initial port; v1 scope = the core check- flow- dedupe chain . 证据：`packages/typescript/goldenpipe/CLAUDE.md`
- 其余 20 条证据见 `AI_CONTEXT_PACK.json` 或 `EVIDENCE_INDEX.json`。

## 宿主 AI 必须遵守的规则

- **把本资产当作开工前上下文，而不是运行环境。**：AI Context Pack 只包含证据化项目理解，不包含目标项目的可执行状态。 证据：`docs/adr/README.md`, `docs/superpowers/CLAUDE.md`, `CLAUDE.md`
- **回答用户时区分可预览内容与必须安装后才能验证的内容。**：安装前体验的消费者价值来自降低误装和误判，而不是伪装成真实运行。 证据：`docs/adr/README.md`, `docs/superpowers/CLAUDE.md`, `CLAUDE.md`

## 用户开工前应该回答的问题

- 你准备在哪个宿主 AI 或本地环境中使用它？
- 你只是想先体验工作流，还是准备真实安装？
- 你最在意的是安装成本、输出质量、还是和现有规则的冲突？

## 验收标准

- 所有能力声明都能回指到 evidence_refs 中的文件路径。
- AI_CONTEXT_PACK.md 没有把预览包装成真实运行。
- 用户能在 3 分钟内看懂适合谁、能做什么、如何开始和风险边界。

---

## Doramagic Context Augmentation

下面内容用于强化 Repomix/AI Context Pack 主体。Human Manual 只提供阅读骨架；踩坑日志会被转成宿主 AI 必须遵守的工作约束。

## Human Manual 骨架

使用规则：这里只是项目阅读路线和显著性信号，不是事实权威。具体事实仍必须回到 repo evidence / Claim Graph。

宿主 AI 硬性规则：
- 不得把页标题、章节顺序、摘要或 importance 当作项目事实证据。
- 解释 Human Manual 骨架时，必须明确说它只是阅读路线/显著性信号。
- 能力、安装、兼容性、运行状态和风险判断必须引用 repo evidence、source path 或 Claim Graph。

- **Home**：importance `high`
  - source_paths: README.md, packages/python/goldenmatch/README.md
- **Getting Started**：importance `high`
  - source_paths: examples/python/01_quickstart_dedupe.py, examples/python/02_full_suite_pipeline.py, packages/python/goldenmatch/examples/basic_dedupe.py
- **Suite Packages Overview**：importance `high`
  - source_paths: packages/python/goldencheck/README.md, packages/python/goldenflow/README.md, packages/python/goldenpipe/README.md, packages/python/infermap/README.md, packages/rust/extensions/README.md
- **Installation**：importance `high`
  - source_paths: packages/python/goldenmatch/pyproject.toml, packages/python/goldensuite-mcp/README.md
- **System Architecture**：importance `high`
  - source_paths: packages/python/goldenmatch/docs/architecture.md, packages/python/goldenmatch/goldenmatch/core/pipeline.py, packages/python/goldenmatch/goldenmatch/core/memory/learner.py
- **Backend Systems**：importance `high`
  - source_paths: packages/python/goldenmatch/goldenmatch/backends/duckdb_backend.py, packages/python/goldenmatch/goldenmatch/backends/ray_backend.py, docs/scale-envelope.md
- **Core Matching Engine**：importance `high`
  - source_paths: packages/python/goldenmatch/goldenmatch/core/scorer.py, packages/python/goldenmatch/goldenmatch/core/cluster.py, packages/python/goldenmatch/goldenmatch/core/autoconfig_controller.py
- **AutoConfig System**：importance `high`
  - source_paths: packages/python/goldenmatch/goldenmatch/core/autoconfig_controller.py, packages/python/goldenmatch/goldenmatch/core/autoconfig_policy.py, packages/python/goldenmatch/goldenmatch/core/autoconfig_cluster_threshold_tuner.py, packages/python/goldenmatch/goldenmatch/core/zero_label_confidence.py, packages/python/goldenmatch/docs/design/2026-05-25-zero-label-confidence-autoconfig-design.md

## Repo Inspection Evidence / 源码检查证据

- repo_clone_verified: true
- repo_inspection_verified: true
- repo_commit: `cbaceee4a5b027e888243b8b34f587265997d838`
- inspected_files: `pyproject.toml`, `pnpm-lock.yaml`, `package.json`, `README.md`, `uv.lock`, `docs/explicit-config.md`, `docs/er-vendor-comparison.md`, `docs/future-work.md`, `docs/scale-audit-2026-05.md`, `docs/org-transfer-2026-05-15.md`, `docs/scale-envelope.md`, `docs/distributed-ray-cluster-setup.md`, `docs/ci-lanes.md`, `docs/duckdb-sql-scoring-research.md`, `docs/reproducing-benchmarks.md`, `docs/distributed-ray-roadmap.md`, `docs/distributed/ray-gce-cluster.md`, `docs/adr/0002-unified-exclude-columns.md`, `docs/adr/0004-chao1-sample-correction.md`, `docs/adr/0000-adopt-adr-practice.md`

宿主 AI 硬性规则：
- 没有 repo_clone_verified=true 时，不得声称已经读过源码。
- 没有 repo_inspection_verified=true 时，不得把 README/docs/package 文件判断写成事实。
- 没有 quick_start_verified=true 时，不得声称 Quick Start 已跑通。

## Doramagic Pitfall Constraints / 踩坑约束

这些规则来自 Doramagic 发现、验证或编译过程中的项目专属坑点。宿主 AI 必须把它们当作工作约束，而不是普通说明文字。

### Constraint 1: 能力判断依赖假设

- Trigger: README/documentation is current enough for a first validation pass.
- Host AI rule: 将假设转成下游验证清单。
- Why it matters: 假设不成立时，用户拿不到承诺的能力。
- Evidence: capability.assumptions | github_repo:1183640892 | https://github.com/benseverndev-oss/goldenmatch | README/documentation is current enough for a first validation pass.
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 2: 维护活跃度未知

- Trigger: 未记录 last_activity_observed。
- Host AI rule: 补 GitHub 最近 commit、release、issue/PR 响应信号。
- Why it matters: 新项目、停更项目和活跃项目会被混在一起，推荐信任度下降。
- Evidence: evidence.maintainer_signals | github_repo:1183640892 | https://github.com/benseverndev-oss/goldenmatch | last_activity_observed missing
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 3: 下游验证发现风险项

- Trigger: no_demo
- Host AI rule: 进入安全/权限治理复核队列。
- Why it matters: 下游已经要求复核，不能在页面中弱化。
- Evidence: downstream_validation.risk_items | github_repo:1183640892 | https://github.com/benseverndev-oss/goldenmatch | no_demo; severity=medium
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 4: 存在评分风险

- Trigger: no_demo
- Host AI rule: 把风险写入边界卡，并确认是否需要人工复核。
- Why it matters: 风险会影响是否适合普通用户安装。
- Evidence: risks.scoring_risks | github_repo:1183640892 | https://github.com/benseverndev-oss/goldenmatch | no_demo; severity=medium
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 5: issue/PR 响应质量未知

- Trigger: issue_or_pr_quality=unknown。
- Host AI rule: 抽样最近 issue/PR，判断是否长期无人处理。
- Why it matters: 用户无法判断遇到问题后是否有人维护。
- Evidence: evidence.maintainer_signals | github_repo:1183640892 | https://github.com/benseverndev-oss/goldenmatch | issue_or_pr_quality=unknown
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 6: 发布节奏不明确

- Trigger: release_recency=unknown。
- Host AI rule: 确认最近 release/tag 和 README 安装命令是否一致。
- Why it matters: 安装命令和文档可能落后于代码，用户踩坑概率升高。
- Evidence: evidence.maintainer_signals | github_repo:1183640892 | https://github.com/benseverndev-oss/goldenmatch | release_recency=unknown
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。