# upsonic - 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 upsonic 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_0003` supported 0.86
- **希望把专业流程带进宿主 AI 的用户**：仓库包含 Skill 文档。 证据：`prebuilt_autonomous_agents/applied_scientist/skills/analyze_current/SKILL.md`, `prebuilt_autonomous_agents/applied_scientist/skills/benchmark/SKILL.md`, `prebuilt_autonomous_agents/applied_scientist/skills/evaluate/SKILL.md`, `prebuilt_autonomous_agents/applied_scientist/skills/experiment_management/SKILL.md` 等 Claim：`clm_0004` supported 0.86

## 它能做什么

- **AI Skill / Agent 指令资产库**（可做安装前预览）：项目包含可被宿主 AI 读取的 Skill 或 Agent 指令文件，可用于把专业流程带入 Claude、Codex、Cursor 等宿主。 证据：`prebuilt_autonomous_agents/applied_scientist/skills/analyze_current/SKILL.md`, `prebuilt_autonomous_agents/applied_scientist/skills/benchmark/SKILL.md`, `prebuilt_autonomous_agents/applied_scientist/skills/evaluate/SKILL.md`, `prebuilt_autonomous_agents/applied_scientist/skills/experiment_management/SKILL.md` 等 Claim：`clm_0001` supported 0.86
- **命令行启动或安装流程**（需要安装后验证）：项目文档中存在可执行命令，真实使用需要在本地或宿主环境中运行这些命令。 证据：`CONTRIBUTING.md`, `documents/ai/explanation/vectordb/vectordb.md` Claim：`clm_0002` supported 0.86

## 怎么开始

- `pip install "upsonic[chroma]"      # chromadb` 证据：`documents/ai/explanation/vectordb/vectordb.md` Claim：`clm_0005` unverified 0.25
- `pip install "upsonic[faiss]"       # faiss-cpu, numpy` 证据：`documents/ai/explanation/vectordb/vectordb.md` Claim：`clm_0006` unverified 0.25
- `pip install "upsonic[qdrant]"      # qdrant-client` 证据：`documents/ai/explanation/vectordb/vectordb.md` Claim：`clm_0007` unverified 0.25
- `pip install "upsonic[pinecone]"    # pinecone, pinecone-text` 证据：`documents/ai/explanation/vectordb/vectordb.md` Claim：`clm_0008` unverified 0.25
- `pip install "upsonic[milvus]"      # pymilvus` 证据：`documents/ai/explanation/vectordb/vectordb.md` Claim：`clm_0009` unverified 0.25
- `pip install "upsonic[weaviate]"    # weaviate-client` 证据：`documents/ai/explanation/vectordb/vectordb.md` Claim：`clm_0010` unverified 0.25
- `pip install "upsonic[pgvector]"    # sqlalchemy, psycopg, pgvector` 证据：`documents/ai/explanation/vectordb/vectordb.md` Claim：`clm_0011` unverified 0.25
- `pip install "upsonic[supermemory]" # supermemory` 证据：`documents/ai/explanation/vectordb/vectordb.md` Claim：`clm_0012` unverified 0.25
- `pip install uv` 证据：`CONTRIBUTING.md` Claim：`clm_0013` 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**：研究结论、引用和实验结果不能在安装前相信。
- **继续会触碰**：研究判断、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_0003` supported 0.86
- **适合人群线索：希望把专业流程带进宿主 AI 的用户**（supported）：有 supported claim 或项目证据支撑，但仍不等于真实安装效果。 证据：`prebuilt_autonomous_agents/applied_scientist/skills/analyze_current/SKILL.md`, `prebuilt_autonomous_agents/applied_scientist/skills/benchmark/SKILL.md`, `prebuilt_autonomous_agents/applied_scientist/skills/evaluate/SKILL.md`, `prebuilt_autonomous_agents/applied_scientist/skills/experiment_management/SKILL.md` 等 Claim：`clm_0004` supported 0.86
- **能力存在：AI Skill / Agent 指令资产库**（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：`prebuilt_autonomous_agents/applied_scientist/skills/analyze_current/SKILL.md`, `prebuilt_autonomous_agents/applied_scientist/skills/benchmark/SKILL.md`, `prebuilt_autonomous_agents/applied_scientist/skills/evaluate/SKILL.md`, `prebuilt_autonomous_agents/applied_scientist/skills/experiment_management/SKILL.md` 等 Claim：`clm_0001` supported 0.86
- **能力存在：命令行启动或安装流程**（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：`CONTRIBUTING.md`, `documents/ai/explanation/vectordb/vectordb.md` Claim：`clm_0002` supported 0.86
- **存在 Quick Start / 安装命令线索**（supported）：可以相信项目文档出现过启动或安装入口；不要因此直接在主力环境运行。 证据：`CONTRIBUTING.md` Claim：`clm_0013` supported 0.86

### 现在还不能相信

- **研究结论、引用和实验结果不能在安装前相信。**（unverified）：研究 Skill 可以组织问题和路径，但不能替代真实资料检索、论文核验和实验复现。
- **是否适合你的具体研究领域不能直接相信。**（unverified）：Skill 覆盖很多研究主题，不代表对你的领域、资料要求和可信度标准足够。
- **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`, `prebuilt_autonomous_agents/applied_scientist/skills/analyze_current/SKILL.md`, `prebuilt_autonomous_agents/applied_scientist/skills/benchmark/SKILL.md`, `prebuilt_autonomous_agents/applied_scientist/skills/evaluate/SKILL.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

- **研究判断**：问题拆解、资料路径、实验路径、结论结构和可信度判断。 原因：研究型 Skill 可能让输出看起来更专业，但不能替代真实证据核验。
- **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：`CONTRIBUTING.md`, `documents/ai/explanation/vectordb/vectordb.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`, `prebuilt_autonomous_agents/applied_scientist/skills/analyze_current/SKILL.md`, `prebuilt_autonomous_agents/applied_scientist/skills/benchmark/SKILL.md`, `prebuilt_autonomous_agents/applied_scientist/skills/evaluate/SKILL.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：`CONTRIBUTING.md`, `documents/ai/explanation/vectordb/vectordb.md`
- **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`, `benchmarks/README.md`, `benchmarks/SETUP.md`, `benchmarks/overhead_analysis/README.md` 等
- **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**：先验证它能否正确界定研究问题和证据边界，不要先相信研究输出。（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。
- **保留资料和结论核验清单**：如果后续发现引用或实验路径不可靠，可以回到证据边界阶段重新校验。
- **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_0014` 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：`CONTRIBUTING.md`, `documents/ai/explanation/vectordb/vectordb.md` Claim：`clm_0015` supported 0.86
- **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

- **AI Skill / Agent 指令资产库**：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：`prebuilt_autonomous_agents/applied_scientist/skills/analyze_current/SKILL.md`, `prebuilt_autonomous_agents/applied_scientist/skills/benchmark/SKILL.md`, `prebuilt_autonomous_agents/applied_scientist/skills/evaluate/SKILL.md`, `prebuilt_autonomous_agents/applied_scientist/skills/experiment_management/SKILL.md` 等 Claim：`clm_0001` supported 0.86
- **命令行启动或安装流程**：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：`CONTRIBUTING.md`, `documents/ai/explanation/vectordb/vectordb.md` Claim：`clm_0002` supported 0.86

### 上下文规模

- 文件总数：633
- 重要文件覆盖：40/633
- 证据索引条目：72
- 角色 / Skill 条目：17

### 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 upsonic 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 upsonic 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 upsonic 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 17 个角色 / Skill / 项目文档条目。

- **Analyze Current Skill**（skill）：Purpose Read and understand the current baseline implementation. Extract all relevant information about the existing approach without modifying anything, and record the analysis as a structured JSON entry. 激活提示：当用户任务与“Analyze Current Skill”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`prebuilt_autonomous_agents/applied_scientist/skills/analyze_current/SKILL.md`
- **Benchmark Skill**（skill）：Purpose Define the comparison metrics and extract baseline values from the current implementation. Record them as a structured JSON entry so downstream phases and final evaluation can read them directly. 激活提示：当用户任务与“Benchmark Skill”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`prebuilt_autonomous_agents/applied_scientist/skills/benchmark/SKILL.md`
- **Evaluate Skill**（skill）：Purpose Compare baseline and new implementation results. Produce the machine-readable final report result.json , update experiments.json , and append a row to comparison.json . 激活提示：当用户任务与“Evaluate Skill”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`prebuilt_autonomous_agents/applied_scientist/skills/evaluate/SKILL.md`
- **Experiment Management Skill**（skill）：Purpose Set up and manage the experiment folder structure. This is Phase 0 — it runs before any analysis begins. All bookkeeping files are JSON never markdown . 激活提示：当用户任务与“Experiment Management Skill”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`prebuilt_autonomous_agents/applied_scientist/skills/experiment_management/SKILL.md`
- **Implement Skill**（skill）：Purpose Create a new Jupyter notebook implementing the method from the research paper, using the same data as the baseline. Record implementation details and measured metrics as a structured JSON entry. 激活提示：当用户任务与“Implement Skill”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`prebuilt_autonomous_agents/applied_scientist/skills/implement/SKILL.md`
- **Progress Skill**（skill）：Purpose Maintain a machine-readable progress file so dashboards, CLIs, and notebooks can poll the experiment's state at any time. The file is a JSON document — never markdown, never human-prose-first. 激活提示：当用户任务与“Progress Skill”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`prebuilt_autonomous_agents/applied_scientist/skills/progress/SKILL.md`
- **Research Skill**（skill）：Purpose Read the materialized research source and extract actionable information needed to implement the proposed method. Record the findings as a structured JSON entry. 激活提示：当用户任务与“Research Skill”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`prebuilt_autonomous_agents/applied_scientist/skills/research/SKILL.md`
- **Analyze Current Skill**（skill）：Purpose Read and understand the current baseline implementation. Extract all relevant information about the existing approach without modifying anything, and record the analysis as a structured JSON entry. 激活提示：当用户任务与“Analyze Current Skill”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`src/upsonic/prebuilt/applied_scientist/template/skills/analyze_current/SKILL.md`
- **Benchmark Skill**（skill）：Purpose Define the comparison metrics and extract baseline values from the current implementation. Record them as a structured JSON entry so downstream phases and final evaluation can read them directly. 激活提示：当用户任务与“Benchmark Skill”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`src/upsonic/prebuilt/applied_scientist/template/skills/benchmark/SKILL.md`
- **Evaluate Skill**（skill）：Purpose Compare baseline and new implementation results. Produce the machine-readable final report result.json , update experiments.json , and append a row to comparison.json . 激活提示：当用户任务与“Evaluate Skill”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`src/upsonic/prebuilt/applied_scientist/template/skills/evaluate/SKILL.md`
- **Experiment Management Skill**（skill）：Purpose Set up and manage the experiment folder structure. This is Phase 0 — it runs before any analysis begins. All bookkeeping files are JSON never markdown . 激活提示：当用户任务与“Experiment Management Skill”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`src/upsonic/prebuilt/applied_scientist/template/skills/experiment_management/SKILL.md`
- **Implement Skill**（skill）：Purpose Create a new Jupyter notebook implementing the method from the research paper, using the same data as the baseline. Record implementation details and measured metrics as a structured JSON entry. 激活提示：当用户任务与“Implement Skill”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`src/upsonic/prebuilt/applied_scientist/template/skills/implement/SKILL.md`
- **Progress Skill**（skill）：Purpose Maintain a machine-readable progress file so dashboards, CLIs, and notebooks can poll the experiment's state at any time. The file is a JSON document — never markdown, never human-prose-first. 激活提示：当用户任务与“Progress Skill”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`src/upsonic/prebuilt/applied_scientist/template/skills/progress/SKILL.md`
- **Research Skill**（skill）：Purpose Read the materialized research source and extract actionable information needed to implement the proposed method. Record the findings as a structured JSON entry. 激活提示：当用户任务与“Research Skill”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`src/upsonic/prebuilt/applied_scientist/template/skills/research/SKILL.md`
- **code-review**（skill）：Perform structured code reviews with actionable feedback. Use when a user asks to review code, check code quality, find bugs, audit security, improve performance, or assess maintainability. Trigger when user says things like "review this code", "check for bugs", "is this code secure", "any issues with this", "code quality check", or pastes code asking for feedback. Also trigger for pull request reviews and pre-merge… 激活提示：当用户任务与“code-review”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`src/upsonic/skills/builtins/code-review/SKILL.md`
- **data-analysis**（skill）：Analyze, explore, clean, and visualize datasets with statistical rigor. Use when user asks to analyze data, find patterns, compute statistics, create visualizations, clean messy data, or explore a dataset. Trigger when user says things like "analyze this data", "what trends do you see", "find patterns in", "create a chart", "clean this dataset", "run statistics on", "what does this data tell us", or provides CSV/Exc… 激活提示：当用户任务与“data-analysis”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`src/upsonic/skills/builtins/data-analysis/SKILL.md`
- **summarization**（skill）：Summarize documents, articles, conversations, code, and technical content into concise, accurate summaries. Use when user asks to summarize, condense, create a TL;DR, write an executive summary, extract key points, or distill content. Trigger when user says things like "summarize this", "give me the key points", "TL;DR", "what are the main takeaways", "condense this", "brief me on this", or provides long content ask… 激活提示：当用户任务与“summarization”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`src/upsonic/skills/builtins/summarization/SKILL.md`

## 证据索引

- 共索引 72 条证据。

- **Upsonic**（documentation）：Build Autonomous AI Agents in Python 证据：`README.md`
- **Upsonic Framework Benchmarks**（documentation）：This directory contains benchmark projects for measuring the performance of various Upsonic Framework components. 证据：`benchmarks/README.md`
- **Overhead Analysis: Direct vs Agent**（documentation）：This benchmark measures the performance differences between Upsonic Framework's Direct LLM Call minimal overhead and Agent full-featured approaches. 证据：`benchmarks/overhead_analysis/README.md`
- **CLAUDE.md**（documentation）：This file provides guidance to Claude Code claude.ai/code when working with code in this repository. 证据：`CLAUDE.md`
- **Contributing to Upsonic**（documentation）：Upsonic is an open-source AI Agent Framework. We welcome contributions that align with our standards. 证据：`CONTRIBUTING.md`
- **Analyze Current Skill**（skill_instruction）：Purpose Read and understand the current baseline implementation. Extract all relevant information about the existing approach without modifying anything, and record the analysis as a structured JSON entry. 证据：`prebuilt_autonomous_agents/applied_scientist/skills/analyze_current/SKILL.md`
- **Benchmark Skill**（skill_instruction）：Purpose Define the comparison metrics and extract baseline values from the current implementation. Record them as a structured JSON entry so downstream phases and final evaluation can read them directly. 证据：`prebuilt_autonomous_agents/applied_scientist/skills/benchmark/SKILL.md`
- **Evaluate Skill**（skill_instruction）：Purpose Compare baseline and new implementation results. Produce the machine-readable final report result.json , update experiments.json , and append a row to comparison.json . 证据：`prebuilt_autonomous_agents/applied_scientist/skills/evaluate/SKILL.md`
- **Experiment Management Skill**（skill_instruction）：Purpose Set up and manage the experiment folder structure. This is Phase 0 — it runs before any analysis begins. All bookkeeping files are JSON never markdown . 证据：`prebuilt_autonomous_agents/applied_scientist/skills/experiment_management/SKILL.md`
- **Implement Skill**（skill_instruction）：Purpose Create a new Jupyter notebook implementing the method from the research paper, using the same data as the baseline. Record implementation details and measured metrics as a structured JSON entry. 证据：`prebuilt_autonomous_agents/applied_scientist/skills/implement/SKILL.md`
- **Progress Skill**（skill_instruction）：Purpose Maintain a machine-readable progress file so dashboards, CLIs, and notebooks can poll the experiment's state at any time. The file is a JSON document — never markdown, never human-prose-first. 证据：`prebuilt_autonomous_agents/applied_scientist/skills/progress/SKILL.md`
- **Research Skill**（skill_instruction）：Purpose Read the materialized research source and extract actionable information needed to implement the proposed method. Record the findings as a structured JSON entry. 证据：`prebuilt_autonomous_agents/applied_scientist/skills/research/SKILL.md`
- **Analyze Current Skill**（skill_instruction）：Purpose Read and understand the current baseline implementation. Extract all relevant information about the existing approach without modifying anything, and record the analysis as a structured JSON entry. 证据：`src/upsonic/prebuilt/applied_scientist/template/skills/analyze_current/SKILL.md`
- **Benchmark Skill**（skill_instruction）：Purpose Define the comparison metrics and extract baseline values from the current implementation. Record them as a structured JSON entry so downstream phases and final evaluation can read them directly. 证据：`src/upsonic/prebuilt/applied_scientist/template/skills/benchmark/SKILL.md`
- **Evaluate Skill**（skill_instruction）：Purpose Compare baseline and new implementation results. Produce the machine-readable final report result.json , update experiments.json , and append a row to comparison.json . 证据：`src/upsonic/prebuilt/applied_scientist/template/skills/evaluate/SKILL.md`
- **Experiment Management Skill**（skill_instruction）：Purpose Set up and manage the experiment folder structure. This is Phase 0 — it runs before any analysis begins. All bookkeeping files are JSON never markdown . 证据：`src/upsonic/prebuilt/applied_scientist/template/skills/experiment_management/SKILL.md`
- **Implement Skill**（skill_instruction）：Purpose Create a new Jupyter notebook implementing the method from the research paper, using the same data as the baseline. Record implementation details and measured metrics as a structured JSON entry. 证据：`src/upsonic/prebuilt/applied_scientist/template/skills/implement/SKILL.md`
- **Progress Skill**（skill_instruction）：Purpose Maintain a machine-readable progress file so dashboards, CLIs, and notebooks can poll the experiment's state at any time. The file is a JSON document — never markdown, never human-prose-first. 证据：`src/upsonic/prebuilt/applied_scientist/template/skills/progress/SKILL.md`
- **Research Skill**（skill_instruction）：Purpose Read the materialized research source and extract actionable information needed to implement the proposed method. Record the findings as a structured JSON entry. 证据：`src/upsonic/prebuilt/applied_scientist/template/skills/research/SKILL.md`
- **Code Review**（skill_instruction）：Perform a structured, multi-dimensional code review. Act as a senior engineer reviewing a colleague's work — be thorough but constructive. 证据：`src/upsonic/skills/builtins/code-review/SKILL.md`
- **Data Analysis**（skill_instruction）：Explore, clean, analyze, and communicate findings from data. The goal is always to answer a question — start with what the user wants to know and work backward to the analysis that answers it. 证据：`src/upsonic/skills/builtins/data-analysis/SKILL.md`
- **Summarization**（skill_instruction）：Produce concise, accurate summaries that capture what matters. A good summary saves the reader time while preserving the information they need to make decisions or take action. 证据：`src/upsonic/skills/builtins/summarization/SKILL.md`
- **Init**（source_file）：version = "0.1.0" ⋮---- all = 证据：`benchmarks/__init__.py`
- **Init**（source_file）：version = "0.1.0" ⋮---- all = "TestCases" 证据：`benchmarks/overhead_analysis/__init__.py`
- **Init**（source_file）：version = "0.77.3" ⋮---- lazy imports = {} cwd = Path os.getcwd env path = cwd / ".env" ⋮---- def lazy import module name: str, class name: str None = None ⋮---- def import ⋮---- def get Task ⋮---- task cls = lazy import "upsonic.tasks.tasks", "Task" ⋮---- def get KnowledgeBase def get Agent ⋮---- agent cls = lazy import "upsonic.agent.agent", "Agent" ⋮---- def get Clanker ⋮---- clanker cls = lazy import "upsonic.agent.agent", "Clanker" ⋮---- def get Graph def get Team def get Chat def get Direct def get Simulation def get RalphLoop def get AutonomousAgent def get PrebuiltAutonomousAgentBase def hello - str def getattr name: str - Any all = 证据：`src/upsonic/__init__.py`
- **Output**（source_file）：@dataclass kw only=True class OutputSchema ABC, Generic OutputDataT ⋮---- text processor: BaseOutputProcessor OutputDataT None = None toolset: None = None object def: OutputObjectDefinition None = None allows deferred tools: bool = False allows image: bool = False ⋮---- @property def mode self - OutputMode ⋮---- @property def allows text self - bool ⋮---- @dataclass init=False class StructuredTextOutputSchema OutputSchema OutputDataT , ABC ⋮---- processor: BaseObjectOutputProcessor OutputDataT template: str None ⋮---- @classmethod def build instructions cls, template: str, object def: OutputObjectDefinition - str ⋮---- schema = object def.json schema.copy ⋮---- template = '\n\n'.join templa… 证据：`src/upsonic/_output.py`
- **Init**（source_file）：def get agent classes def get autonomous agent classes def get event classes def getattr name: str - Any ⋮---- agent classes = get agent classes ⋮---- autonomous classes = get autonomous agent classes ⋮---- event classes = get event classes ⋮---- all = 证据：`src/upsonic/agent/__init__.py`
- **Init**（source_file）：def get classes - dict str, Any def getattr name: str - Any ⋮---- classes = get classes ⋮---- all = 证据：`src/upsonic/agent/autonomous_agent/__init__.py`
- **Base**（source_file）：class BaseAgent ABC 证据：`src/upsonic/agent/base.py`
- **Init**（source_file）：def get context manager classes def getattr name: str - Any ⋮---- context manager classes = get context manager classes ⋮---- all = 证据：`src/upsonic/agent/context_managers/__init__.py`
- **Backends**（source_file）：def get backend classes def get tool classes def get deepagent class def getattr name: str - Any ⋮---- backend classes = get backend classes ⋮---- tool classes = get tool classes ⋮---- deepagent class = get deepagent class ⋮---- all = ⋮---- Backends 证据：`src/upsonic/agent/deepagent/__init__.py`
- **Init**（source_file）：def get backend classes def getattr name: str - Any ⋮---- backend classes = get backend classes ⋮---- all = 证据：`src/upsonic/agent/deepagent/backends/__init__.py`
- **Init**（source_file）：def get toolkit classes def getattr name: str - Any ⋮---- toolkit classes = get toolkit classes ⋮---- all = 证据：`src/upsonic/agent/deepagent/tools/__init__.py`
- **Events**（source_file）：all = 证据：`src/upsonic/agent/events.py`
- **Init**（source_file）：def get pipeline base classes def get pipeline step classes def getattr name: str - Any ⋮---- base classes = get pipeline base classes ⋮---- step classes = get pipeline step classes ⋮---- all = 证据：`src/upsonic/agent/pipeline/__init__.py`
- **Feedback loop support**（source_file）：@dataclass class PolicyScope ⋮---- description: bool context: bool system prompt: bool chat history: bool tool outputs: bool ⋮---- def resolve policy attr: str, task attr: str, agent attr: str - bool ⋮---- val = getattr policy, policy attr, None ⋮---- val = getattr task, task attr, None ⋮---- class PolicyResult ⋮---- def init self ⋮---- Feedback loop support self.feedback message: Optional str = None The feedback if generated self.requires retry: bool = False Whether agent should retry with feedback self.original content: Optional str = None Content that violated policy self.violated policy name: Optional str = None Name of the first violated policy self.violation reason: Optional str = Non… 证据：`src/upsonic/agent/policy_manager.py`
- **DisallowedOperation - stop immediately**（source_file）：class ToolPolicyResult ⋮---- def init self def should block self - bool ⋮---- """Check if tool should be blocked.""" ⋮---- def get final message self - str ⋮---- """Get the final message to return.""" ⋮---- class ToolPolicyManager ⋮---- def has policies self - bool ⋮---- result = ToolPolicyResult ⋮---- policy input = PolicyInput ⋮---- action taken = policy output.action output.get "action taken", "UNKNOWN" ⋮---- DisallowedOperation - stop immediately ⋮---- Create mock rule output for logging ⋮---- mock rule output = RuleOutput ⋮---- Execute policy ⋮---- def setup policy models self, model - None ⋮---- Setup base llm for actions that use LLM ⋮---- def repr self - str ⋮---- """String represen… 证据：`src/upsonic/agent/tool_policy_manager.py`
- **Init**（source_file）：def getattr name: str - Any all = "CacheManager" 证据：`src/upsonic/cache/__init__.py`
- **Init**（source_file）：def getattr name: str - Any all = 'Canvas' 证据：`src/upsonic/canvas/__init__.py`
- **Init**（source_file）：def get chat classes - dict def getattr name: str - Any ⋮---- chat classes = get chat classes ⋮---- all = 证据：`src/upsonic/chat/__init__.py`
- **Init**（source_file）：all = "main" 证据：`src/upsonic/cli/__init__.py`
- **Init**（source_file）：all = 证据：`src/upsonic/cli/commands/__init__.py`
- **Add the dependency**（source_file）：def add command library: str, section: str - int ⋮---- current dir = Path.cwd config json path = current dir / "upsonic configs.json" ⋮---- config data = load config config json path, use cache=False ⋮---- dependencies = config data "dependencies" ⋮---- available sections = list dependencies.keys ⋮---- Add the dependency ⋮---- Write back to file 证据：`src/upsonic/cli/commands/add/command.py`
- **Command**（source_file）：def init command - int ⋮---- agent name = prompt agent name ⋮---- current dir = Path.cwd main py path = current dir / "main.py" config json path = current dir / "upsonic configs.json" ⋮---- main py content = """from upsonic import Task, Agent ⋮---- config data = { 证据：`src/upsonic/cli/commands/init/command.py`
- **Command**（source_file）：def install command section: Optional str = None - int ⋮---- current dir = Path.cwd config json path = current dir / "upsonic configs.json" ⋮---- config data = load config config json path ⋮---- all dependencies = config data.get "dependencies", {} ⋮---- sections to install = "api" ⋮---- sections to install = list all dependencies.keys ⋮---- sections to install = section ⋮---- available sections = list all dependencies.keys ⋮---- dependencies to install = 证据：`src/upsonic/cli/commands/install/command.py`
- **Remove the dependency**（source_file）：def get package name dependency str: str - str ⋮---- delimiters = "==", " =", " ", " int ⋮---- current dir = Path.cwd config json path = current dir / "upsonic configs.json" ⋮---- config data = load config config json path, use cache=False ⋮---- dependencies = config data "dependencies" ⋮---- available sections = list dependencies.keys ⋮---- section deps = dependencies section target lib lower = library.lower target pkg name = get package name library to remove = None ⋮---- to remove = dep ⋮---- Remove the dependency ⋮---- Write back to file 证据：`src/upsonic/cli/commands/remove/command.py`
- **Detect interface mode: static source check then call main**（source_file）：def is interface mode source path: Path - bool ⋮---- source: str = source path.read text encoding="utf-8" ⋮---- def resolve interface manager main func: Callable ..., Any - Optional object ⋮---- result: Any = asyncio.run main func {} ⋮---- result = main func {} ⋮---- def run command host: str = "0.0.0.0", port: int = 8000 - int ⋮---- current dir: Path = Path.cwd config json path: Path = current dir / "upsonic configs.json" ⋮---- config data: Optional dict str, Any = load config config json path ⋮---- agent name: str = config data.get "agent name", "Upsonic Agent" description: str = config data.get "description", "An Upsonic AI agent" entrypoints: dict str, Any = config data.get "entrypoints… 证据：`src/upsonic/cli/commands/run/command.py`
- **Openapi**（source_file）：def map inputs props inputs schema: List Dict str, Any - tuple Dict str, Any , Dict str, Any , List str ⋮---- json props = {} multipart props = {} required = ⋮---- name = item "name" itype = item.get "type", "string" default = item.get "default" ⋮---- def map output props output schema: Dict str, Any - Dict str, Any ⋮---- props = {} ⋮---- t = v.get "type", "string" ⋮---- paths = schema.get "paths", {} ⋮---- post op = paths path .get "post", {} ⋮---- request model = { ⋮---- components = schema.setdefault "components", {} comps schemas = components.setdefault "schemas", {} ⋮---- content = {} multipart schema = { ⋮---- responses = post op.setdefault "responses", {} 证据：`src/upsonic/cli/commands/shared/openapi.py`
- **Collect files to include**（source_file）：def zip command output file: Optional str = None - int ⋮---- current dir = Path.cwd ⋮---- timestamp = datetime.datetime.now .strftime "%Y%m%d %H%M%S" output file = f"upsonic context {timestamp}.zip" ⋮---- output path = current dir / output file ⋮---- Collect files to include files to zip = total size = 0 ⋮---- size = item.stat .st size ⋮---- Create the zip file ⋮---- Add all files ⋮---- relative path = file path.relative to current dir ⋮---- Print success final size = output path.stat .st size 证据：`src/upsonic/cli/commands/zip/command.py`
- **Check for help flag**（source_file）：COMMAND HANDLERS = { def handle init args: list str - int def handle add args: list str - int def handle remove args: list str - int def handle install args: list str - int ⋮---- section = args 1 if len args = 2 else None ⋮---- def handle run args: list str - int ⋮---- host = "0.0.0.0" port = 8000 i = 1 ⋮---- arg = args i ⋮---- host = args i + 1 ⋮---- port = int args i + 1 ⋮---- def handle zip args: list str - int ⋮---- """Handle 'zip' command with lazy import.""" Check for help flag ⋮---- output file = args 1 if len args = 2 else None ⋮---- def main args: Optional list str = None - int ⋮---- args = sys.argv 1: ⋮---- command = args 0 handler = COMMAND HANDLERS.get command 证据：`src/upsonic/cli/main.py`
- **Center the CLI text - UPSONIC is 60 chars wide, CLI is 20 chars, so pad with 16 spaces on left**（source_file）：RICH IMPORTS = None def get rich imports ⋮---- RICH IMPORTS = { ⋮---- def escape rich markup text: str - str ⋮---- rich = get rich imports ⋮---- def print banner - None ⋮---- GREEN BOLD = "\033 1;32m" RESET = "\033 0m" BOLD = "\033 1m" BLUE = "\033 34m" ⋮---- Center the CLI text - UPSONIC is 60 chars wide, CLI is 20 chars, so pad with 16 spaces on left ⋮---- def prompt agent name - str ⋮---- """Prompt user for agent name with styled input.""" ⋮---- console = rich 'console' Prompt = rich 'Prompt' ⋮---- agent name = Prompt.ask " bold Agent Name /bold ", default="" ⋮---- def print error message: str - None ⋮---- """Print an error message in a styled panel.""" ⋮---- Panel = rich 'Panel' box = r… 证据：`src/upsonic/cli/printer.py`
- **Init**（source_file）：def get culture classes def getattr name: str - Any ⋮---- culture classes = get culture classes ⋮---- all = 证据：`src/upsonic/culture/__init__.py`
- **Init**（source_file）：def get database classes def getattr name: str - Any ⋮---- database classes = get database classes ⋮---- all = 证据：`src/upsonic/db/__init__.py`
- **Database**（source_file）：StorageType = TypeVar 'StorageType', bound=Storage class DatabaseBase Generic StorageType ⋮---- @property def session id self - Optional str ⋮---- @property def user id self - Optional str def repr self - str class SqliteDatabase DatabaseBase "SqliteStorage" ⋮---- storage = SqliteStorage memory = Memory ⋮---- class PostgresDatabase DatabaseBase "PostgresStorage" ⋮---- storage = PostgresStorage ⋮---- class MongoDatabase DatabaseBase "MongoStorage" ⋮---- storage = MongoStorage ⋮---- class RedisDatabase DatabaseBase "RedisStorage" ⋮---- storage = RedisStorage ⋮---- class InMemoryDatabase DatabaseBase InMemoryStorage ⋮---- storage = InMemoryStorage ⋮---- class JSONDatabase DatabaseBase JSONStor… 证据：`src/upsonic/db/database.py`
- **Base classes always available**（source_file）：def get base classes def get factory functions def get openai embedding def get openai embedding config def get azure openai embedding def get azure openai embedding config def get bedrock embedding def get bedrock embedding config def get huggingface embedding def get huggingface embedding config def get fastembed provider def get fastembed config def get ollama embedding def get ollama embedding config def get gemini embedding def get gemini embedding config def get gemini vertex embedding def get gemini document embedding def get gemini query embedding def get gemini semantic embedding def get gemini cloud embedding def get azure embedding with managed identity def get create titan embed… 证据：`src/upsonic/embeddings/__init__.py`
- **Base**（source_file）：NUMPY AVAILABLE = True ⋮---- np = None NUMPY AVAILABLE = False ⋮---- class EmbeddingMode str, Enum ⋮---- DOCUMENT = "document" QUERY = "query" SYMMETRIC = "symmetric" CLUSTERING = "clustering" class EmbeddingMetrics BaseModel ⋮---- total chunks: int = 0 total tokens: int = 0 embedding time ms: float = 0 avg time per chunk: float = 0 dimension: int = 0 model name: str = "" provider: str = "" model config = ConfigDict arbitrary types allowed=True class EmbeddingConfig BaseModel ⋮---- """Base configuration for all embedding providers.""" model name: str = Field ..., description="The name of the embedding model" dimension: Optional int = Field None, description="Expected embedding dimension for… 证据：`src/upsonic/embeddings/base.py`
- **Init**（source_file）：def get evaluator classes def get model classes def getattr name: str - Any ⋮---- evaluator classes = get evaluator classes ⋮---- model classes = get model classes ⋮---- all = 证据：`src/upsonic/eval/__init__.py`
- **Init**（source_file）：def get graph classes def getattr name: str - Any ⋮---- graph classes = get graph classes ⋮---- all = 证据：`src/upsonic/graph/__init__.py`
- **Core graph components**（source_file）：def get graphv2 core classes def get graphv2 checkpoint classes def get graphv2 primitives def get graphv2 store classes def get graphv2 cache classes def get graphv2 task classes def get graphv2 error classes def getattr name: str - Any ⋮---- core classes = get graphv2 core classes ⋮---- checkpoint classes = get graphv2 checkpoint classes ⋮---- primitives = get graphv2 primitives ⋮---- store classes = get graphv2 store classes ⋮---- cache classes = get graphv2 cache classes ⋮---- task classes = get graphv2 task classes ⋮---- error classes = get graphv2 error classes ⋮---- all = ⋮---- Core graph components 证据：`src/upsonic/graphv2/__init__.py`
- **Init**（source_file）：def getattr name: str - Any all = "TracingProvider", "DefaultTracingProvider", "Langfuse", "PromptLayer", "AsqavGovernance" 证据：`src/upsonic/integrations/__init__.py`
- 其余 12 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：`README.md`, `benchmarks/README.md`, `benchmarks/overhead_analysis/README.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：`README.md`, `benchmarks/README.md`, `benchmarks/overhead_analysis/README.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。

- **概述、安装与快速入门**：importance `high`
  - source_paths: README.md, pyproject.toml, src/upsonic/__init__.py, src/upsonic/direct.py, src/upsonic/cli/main.py
- **智能体、团队与图编排**：importance `high`
  - source_paths: src/upsonic/agent/agent.py, src/upsonic/agent/autonomous_agent/autonomous_agent.py, src/upsonic/agent/deepagent/deepagent.py, src/upsonic/agent/context_managers/context_manager.py, src/upsonic/agent/pipeline/manager.py
- **工具、MCP 集成与技能系统**：importance `high`
  - source_paths: src/upsonic/tools/base.py, src/upsonic/tools/builtin_tools.py, src/upsonic/tools/mcp.py, src/upsonic/tools/hitl.py, src/upsonic/tools/registry.py
- **数据、存储、知识库与外部接口**：importance `high`
  - source_paths: src/upsonic/storage/base.py, src/upsonic/storage/postgres/postgres.py, src/upsonic/vectordb/factory.py, src/upsonic/vectordb/providers/qdrant.py, src/upsonic/embeddings/factory.py

## Repo Inspection Evidence / 源码检查证据

- repo_clone_verified: true
- repo_inspection_verified: true
- repo_commit: `101f0313b0ddb96cd4078354879b2ff57005db29`
- inspected_files: `README.md`, `pyproject.toml`, `uv.lock`, `src/upsonic/__init__.py`, `src/upsonic/_griffe.py`, `src/upsonic/_json_schema.py`, `src/upsonic/_output.py`, `src/upsonic/_parts_manager.py`, `src/upsonic/_ssrf.py`, `src/upsonic/_utils.py`, `src/upsonic/agent/__init__.py`, `src/upsonic/agent/agent.py`, `src/upsonic/agent/autonomous_agent/__init__.py`, `src/upsonic/agent/autonomous_agent/autonomous_agent.py`, `src/upsonic/agent/autonomous_agent/filesystem_toolkit.py`, `src/upsonic/agent/autonomous_agent/shell_toolkit.py`, `src/upsonic/agent/base.py`, `src/upsonic/agent/context_managers/__init__.py`, `src/upsonic/agent/context_managers/call_manager.py`, `src/upsonic/agent/context_managers/context_management_middleware.py`

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: 来源证据：feat(vectordb): Add Valkey Search as a vector database provider

- Trigger: GitHub 社区证据显示该项目存在一个安装相关的待验证问题：feat(vectordb): Add Valkey Search as a vector database provider
- Why it matters: 可能影响升级、迁移或版本选择。
- Evidence: community_evidence:github | https://github.com/Upsonic/Upsonic/issues/603 | 来源类型 github_issue 暴露的待验证使用条件。
- 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: 可能修改宿主 AI 配置

- Trigger: 项目面向 Claude/Cursor/Codex/Gemini/OpenCode 等宿主，或安装命令涉及用户配置目录。
- Host AI rule: 列出会写入的配置文件、目录和卸载/回滚步骤。
- Why it matters: 安装可能改变本机 AI 工具行为，用户需要知道写入位置和回滚方法。
- Evidence: capability.host_targets | https://github.com/Upsonic/Upsonic | host_targets=mcp_host, claude, openclaw, cursor
- 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: 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 | https://github.com/Upsonic/Upsonic | 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 4: 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 | https://github.com/Upsonic/Upsonic | 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。

- Trigger: no_demo
- Evidence: downstream_validation.risk_items | https://github.com/Upsonic/Upsonic | 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 6: 存在评分风险

- Trigger: no_demo
- Why it matters: 风险会影响是否适合普通用户安装。
- Evidence: risks.scoring_risks | https://github.com/Upsonic/Upsonic | 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 7: 来源证据：Best-practice: SSL context, SQLAlchemy text(f''), agent shell=True comment, and pickle persistence

- Trigger: GitHub 社区证据显示该项目存在一个安全/权限相关的待验证问题：Best-practice: SSL context, SQLAlchemy text(f''), agent shell=True comment, and pickle persistence
- Why it matters: 可能增加新用户试用和生产接入成本。
- Evidence: community_evidence:github | https://github.com/Upsonic/Upsonic/issues/596 | 来源讨论提到 python 相关条件，需在安装/试用前复核。
- 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 8: 来源证据：OwnX Network's hackathon offer for Upsonic

- Trigger: GitHub 社区证据显示该项目存在一个安全/权限相关的待验证问题：OwnX Network's hackathon offer for Upsonic
- Why it matters: 可能影响授权、密钥配置或安全边界。
- Evidence: community_evidence:github | https://github.com/Upsonic/Upsonic/issues/619 | 来源类型 github_issue 暴露的待验证使用条件。
- 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 9: 来源证据：Possible collab: OpenxAI x Upsonic

- Trigger: GitHub 社区证据显示该项目存在一个安全/权限相关的待验证问题：Possible collab: OpenxAI x Upsonic
- Why it matters: 可能影响授权、密钥配置或安全边界。
- Evidence: community_evidence:github | https://github.com/Upsonic/Upsonic/issues/609 | 来源类型 github_issue 暴露的待验证使用条件。
- 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 10: 来源证据：Possible collab: OwnX x Upsonic

- Trigger: GitHub 社区证据显示该项目存在一个安全/权限相关的待验证问题：Possible collab: OwnX x Upsonic
- Why it matters: 可能影响授权、密钥配置或安全边界。
- Evidence: community_evidence:github | https://github.com/Upsonic/Upsonic/issues/609 | 来源类型 github_issue 暴露的待验证使用条件。
- 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。