# mnemoq - 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 mnemoq 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_0002` supported 0.86

## 它能做什么

- **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：`README.md` Claim：`clm_0001` supported 0.86

## 怎么开始

- `pip install mnemoq` 证据：`README.md` Claim：`clm_0003` supported 0.86
- `pipx install mnemoq` 证据：`README.md` Claim：`clm_0004` supported 0.86
- `git clone https://github.com/Mnemoq/MnemoQ.git` 证据：`README.md` Claim：`clm_0005` supported 0.86
- `pip install -e ".[dev]"` 证据：`README.md` Claim：`clm_0006` supported 0.86

## 继续前判断卡

- **当前建议**：先做角色匹配试用
- **为什么**：这个项目更像角色库，核心风险是选错角色或把角色文案当执行能力；先用 Prompt Preview 试角色匹配，再决定是否沙盒导入。

### 30 秒判断

- **现在怎么做**：先做角色匹配试用
- **最小安全下一步**：先用 Prompt Preview 试角色匹配；满意后再隔离导入
- **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_0002` 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：`README.md` Claim：`clm_0001` 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：`README.md` Claim：`clm_0003` 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：`AGENTS.md`
- **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`
- **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：`AGENTS.md`
- **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：`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。）
- **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。
- **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_0007` 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` Claim：`clm_0008` 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

- **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：`README.md` Claim：`clm_0001` supported 0.86

### 上下文规模

- 文件总数：170
- 重要文件覆盖：40/170
- 证据索引条目：77
- 角色 / Skill 条目：59

### 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 mnemoq 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 mnemoq 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 mnemoq 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 59 个角色 / Skill / 项目文档条目。

- **MnemoQ Documentation**（project_doc）：Complete documentation for the MnemoQ agent memory engine. Start here if you're new. 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/README.md`
- **MnemoQ**（project_doc）：Local-first memory engine for AI agents — MCP-native, graph-linked, spaced repetition. 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`
- **Contributing to MnemoQ**（project_doc）：Thank you for your interest in contributing! This document covers everything you need 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：`CONTRIBUTING.md`
- **Agent Guidelines**（project_doc）：Architecture cli.py is a thin dispatcher — all logic lives in src/agent memory/engine/ modules. Always pass ctx dict and Paths to engine functions; never read module globals directly. 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：`AGENTS.md`
- **Roadmap**（project_doc）：- x BM25 lexical scoring + Reciprocal Rank Fusion RRF - x Embedding-based retrieval sentence-transformers , hybrid scoring - x Embedding-based semantic dedup cosine ≥ 0.85 → merge - x Optional reranking pass cross-encoder, LLM-local - x Grading harness --eval 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/ROADMAP.md`
- **Agent Guidelines**（project_doc）：Architecture cli.py is a thin dispatcher — all logic lives in src/agent memory/engine/ modules. Always pass ctx dict and Paths to engine functions; never read module globals directly. 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：`memory/AGENTS.md`
- **Project Instructions**（project_doc）：Session start: memory/HANDOFF.md and memory/SYSTEM INVARIANTS.md are auto-loaded by your IDE/agent platform. Act on HANDOFF's "next action" line if present. 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：`templates/claude-code/CLAUDE.md`
- **Changelog**（project_doc）：All notable changes to MnemoQ will be documented in this 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：`CHANGELOG.md`
- **Architecture Overview**（project_doc）：MnemoQ is a local-first memory engine for AI agents. It stores learnings as JSONL, retrieves them via a multi-channel scoring pipeline, and integrates with any MCP-compatible client. This doc gives newcomers a conceptual map and contributors a module-level reference. 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/architecture-overview.md`
- **CLI Reference**（project_doc）：Complete reference for all MnemoQ command-line tools. 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/cli-reference.md`
- **Config Tuning Guide**（project_doc）：config.json provides project-specific tuning overlaid on engine defaults. Every parameter below has a sensible default in src/agent memory/engine/constants.py ; your memory/config.json overrides only what you need. See templates/config.json for the full template and templates/config-presets/ for ready-made presets. 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/config-tuning.md`
- **Data Schema Reference**（project_doc）：Canonical reference for the learnings.jsonl entry schema. Each line is a JSON object conforming to the fields below. 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/data-schema.md`
- **MCP Integration Guide**（project_doc）：The Model Context Protocol MCP is the primary integration path for AI agents to access MnemoQ's memory engine. This guide covers installation, client configuration, tool reference, and troubleshooting. 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/mcp-integration.md`
- **Open-Core Architecture**（project_doc）：MnemoQ uses an open-core model: the AGPL-3.0-or-later core lives in this public repo, while a proprietary Pro tier runs from a separate private repo. 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/open-core-architecture.md`
- **Python SDK Guide**（project_doc）：Programmatic access to the MnemoQ memory engine — log, retrieve, update, resolve, consolidate, and read metrics from Python code. 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/sdk-guide.md`
- **Contributor License Agreement**（project_doc）：Thank you for your interest in contributing to MnemoQ "the Project" . 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：`CLA.md`
- **Security Policy**（project_doc）：Do not open a public issue for security vulnerabilities. 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：`SECURITY.md`
- **Handoff - 2026-06-25**（project_doc）：Session Summary Update this section at end of each session 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：`memory/HANDOFF.md`
- **System Invariants**（project_doc）：Consolidated structural rules. IMMUTABLE during active tasks. Only updated during Sleep Cycle. 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：`memory/SYSTEM_INVARIANTS.md`
- **Memory**（project_doc）：Session start: memory/HANDOFF.md and memory/SYSTEM INVARIANTS.md are auto-loaded by your IDE/agent platform. Act on HANDOFF's "next action" line if present. 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：`templates/agents-memory-section.md`
- **Pro**（project_doc）： 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：`.devin/workflows/pro.md`
- **Your Mission**（project_doc）：You are the Fuzzer. A feature has just been implemented. Your job is to try and break 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：`.opencode/Prompts/Fuzzer.md`
- **Review Priority Order**（project_doc）：You are a senior code reviewer for this project. You review diffs against this project's engineering rules and produce a structured report with severity-ranked findings. 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：`.opencode/Prompts/code-reviewer.md`
- **Role and Identity**（project_doc）：Role and Identity You are GM , the primary co-developer and orchestrator for this project. You are highly autonomous and strictly action-oriented. You exist to build exceptional software alongside the human developer, keeping their session entirely clean of tool noise. 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：`.opencode/Prompts/gm.md`
- **Data Sources read all before analysis**（project_doc）：You are the Meta-agent. Your sole purpose is to analyze agent performance data and evolve agent prompts to eliminate recurring failures. 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：`.opencode/Prompts/meta-agent.md`
- **Advanced Metrics & Analytics System**（project_doc）：Advanced Metrics & Analytics System 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：`.windsurf/Plans/advanced-metrics-system-686f76.md`
- **Auto-Learning System**（project_doc）：Plan ID: cb6d42 — saved in-repo for consistency with other plans. Audit pass 1d2cfd applied: reworked the retrieval filter data-loss fix , check staleness ctx guard, consolidate gating, API cache/alert hooks, and accuracy notes. See the audit report for 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：`.windsurf/Plans/auto-learning-system-cb6d42.md`
- **Fake Memory Generator**（project_doc）：A script to bulk-generate valid synthetic memory entries for stress-testing the Agent Memory Engine pipeline, with direct-file and full-pipeline modes. 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：`.windsurf/Plans/fake-memory-generator-8e535e.md`
- **Memory Engine GUI — Web Dashboard, Tauri Desktop, TUI**（project_doc）：Memory Engine GUI — Web Dashboard, Tauri Desktop, TUI 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：`.windsurf/Plans/memory-engine-gui-686f76.md`
- **Agent Memory Engine — Consolidated Roadmap**（project_doc）：Agent Memory Engine — Consolidated Roadmap 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：`.windsurf/Plans/memory-engine-roadmap-consolidated.md`
- **Memory Upgrade Progression — Sequential Implementation Guide**（project_doc）：Memory Upgrade Progression — Sequential Implementation Guide 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：`.windsurf/Plans/memory-upgrade-progression.md`
- **Tier 1 — Free Tier Quality Foundation v1.17 – v1.19**（project_doc）：Tier 1 — Free Tier Quality Foundation v1.17 – v1.19 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：`.windsurf/Plans/tier-1-expanded.md`
- **Tier 2 — Distribution & Access v1.20 – v1.21**（project_doc）：Tier 2 — Distribution & Access v1.20 – v1.21 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：`.windsurf/Plans/tier-2-expanded.md`
- **Web Dashboard 2.3 — Audit Follow-Up Plan**（project_doc）：Web Dashboard 2.3 — Audit Follow-Up Plan 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：`.windsurf/Plans/web-dashboard-2.3-audit-follow-up.md`
- **Web Dashboard 2.3 — Audit Report**（project_doc）：Date: 2026-06-24 Scope: Dashboard backend src/engine/server.py , dashboard api.py , analysis.py , metrics.py and frontend src/dashboard/static/ Test run: python -m pytest tests/ -q → 125 passed 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：`.windsurf/Plans/web-dashboard-2.3-audit-report.md`
- **Web Dashboard 2.3 — Three-Phase Implementation Plan**（project_doc）：Web Dashboard 2.3 — Three-Phase Implementation Plan 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：`.windsurf/Plans/web-dashboard-2.3.md`
- **Review Priority Order**（project_doc）：Reviews code diffs against project rules. Structured report with severity-ranked findings. Read-only. 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：`.windsurf/workflows/code-reviewer.md`
- **Your Mission**（project_doc）：Keeps READMEs, API docs, and inline comments in sync with code changes. Only touches .md files. 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：`.windsurf/workflows/docs-writer.md`
- **Your Mission**（project_doc）：Context gatherer — maps how a feature works across the codebase, returns focused summary. Read-only. 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：`.windsurf/workflows/explorer.md`
- **Your Mission**（project_doc）：Adversarial tester — writes edge-case tests, runs them, reports failures. Never edits src/. 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：`.windsurf/workflows/fuzzer.md`
- **Core Directives**（project_doc）：Primary co-developer and orchestrator. Drives tasks to completion, verifies, and commits. 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：`.windsurf/workflows/gm.md`
- **Environment**（project_doc）：Meta-agent that analyzes failure patterns from recent sessions and evolves subagent prompts to eliminate recurring failures. 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：`.windsurf/workflows/meta-agent.md`
- **Plan Deviation Protocol**（project_doc）：Surface plan deviations as decision points before implementing 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：`.windsurf/workflows/plan-deviation.md`
- **Readiness Rubric**（project_doc）：Audits plan files for readiness. Scores 0-5, identifies gaps, consolidates clarifying questions. Read-only. 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：`.windsurf/workflows/plan-reviewer.md`
- **Steps**（project_doc）：Stages changes cleanly and produces a well-structured Conventional Commit message with branch hygiene checks. 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：`.windsurf/workflows/professional-commit.md`
- **Your Mission**（project_doc）：Structural changes extract, rename, split without changing behavior. Runs tests after each step. 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：`.windsurf/workflows/refactorer.md`
- **Your Mission**（project_doc）：Security auditor — hardcoded secrets, injection, missing auth, unsafe deserialization. Read-only. 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：`.windsurf/workflows/security.md`
- **Agent Guidelines**（project_doc）：Physics Collision detection uses AABB broadphase. Always check penetration depth. 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：`memory/.baseline/fixture-agents.md`
- **Project Instructions**（project_doc）：Session start: memory/HANDOFF.md and memory/SYSTEM INVARIANTS.md are auto-loaded by your IDE/agent platform. Act on HANDOFF's "next action" line if present. 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：`templates/copilot/copilot-instructions.md`
- **Review Priority Order**（project_doc）：You are a senior code reviewer for this project. You review diffs against this project's engineering rules and produce a structured report with severity-ranked findings. 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：`templates/prompts/code-reviewer.md`
- **Your Mission**（project_doc）：You are the Docs Writer. Your job is to keep READMEs, API docs, and inline comments in sync with code changes. 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：`templates/prompts/docs-writer.md`
- **Your Mission**（project_doc）：You are the Explorer. Your job is to map how a feature works across the codebase and return a focused, structured summary. You are cheap, high-volume reading — the parent agent plans against your summary. 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：`templates/prompts/explorer.md`
- **Your Mission**（project_doc）：You are the Fuzzer. A feature has just been implemented. Your job is to try and break 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：`templates/prompts/fuzzer.md`
- **Role and Identity**（project_doc）：Role and Identity You are GM , the primary co-developer and orchestrator for this project. You are highly autonomous and strictly action-oriented. You exist to build exceptional software alongside the human developer, keeping their session entirely clean of tool noise. 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：`templates/prompts/gm.md`
- **Data Sources read all before analysis**（project_doc）：You are the Meta-agent. Your sole purpose is to analyze agent performance data and evolve agent prompts to eliminate recurring failures. 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：`templates/prompts/meta-agent.md`
- **Your Mission**（project_doc）：You are the Refactorer. Your job is to make structural changes extract functions, rename across files, split modules without changing behavior. 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：`templates/prompts/refactorer.md`
- **Your Mission**（project_doc）：You are the Security Auditor. Your job is to perform a focused security pass on the codebase, looking for hardcoded secrets, injection-prone queries, missing authorization, unsafe deserialization, and risky dependencies. 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：`templates/prompts/security.md`
- **Review Priority Order**（project_doc）：Reviews code diffs against project rules. Structured report with severity-ranked findings. Read-only. 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：`templates/windsurf/workflows/code-reviewer.md`
- **Your Mission**（project_doc）：Keeps READMEs, API docs, and inline comments in sync with code changes. Only touches .md files. 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：`templates/windsurf/workflows/docs-writer.md`

## 证据索引

- 共索引 77 条证据。

- **MnemoQ Documentation**（documentation）：Complete documentation for the MnemoQ agent memory engine. Start here if you're new. 证据：`docs/README.md`
- **MnemoQ**（documentation）：Local-first memory engine for AI agents — MCP-native, graph-linked, spaced repetition. 证据：`README.md`
- **Contributing to MnemoQ**（documentation）：Thank you for your interest in contributing! This document covers everything you need to get started. 证据：`CONTRIBUTING.md`
- **Agent Guidelines**（documentation）：Architecture cli.py is a thin dispatcher — all logic lives in src/agent memory/engine/ modules. Always pass ctx dict and Paths to engine functions; never read module globals directly. 证据：`AGENTS.md`
- **Roadmap**（documentation）：- x BM25 lexical scoring + Reciprocal Rank Fusion RRF - x Embedding-based retrieval sentence-transformers , hybrid scoring - x Embedding-based semantic dedup cosine ≥ 0.85 → merge - x Optional reranking pass cross-encoder, LLM-local - x Grading harness --eval 证据：`docs/ROADMAP.md`
- **Agent Guidelines**（documentation）：Architecture cli.py is a thin dispatcher — all logic lives in src/agent memory/engine/ modules. Always pass ctx dict and Paths to engine functions; never read module globals directly. 证据：`memory/AGENTS.md`
- **Project Instructions**（documentation）：Session start: memory/HANDOFF.md and memory/SYSTEM INVARIANTS.md are auto-loaded by your IDE/agent platform. Act on HANDOFF's "next action" line if present. 证据：`templates/claude-code/CLAUDE.md`
- **License**（source_file）：GNU AFFERO GENERAL PUBLIC LICENSE Version 3, 19 November 2007 证据：`LICENSE`
- **Changelog**（documentation）：All notable changes to MnemoQ will be documented in this file. 证据：`CHANGELOG.md`
- **Architecture Overview**（documentation）：MnemoQ is a local-first memory engine for AI agents. It stores learnings as JSONL, retrieves them via a multi-channel scoring pipeline, and integrates with any MCP-compatible client. This doc gives newcomers a conceptual map and contributors a module-level reference. 证据：`docs/architecture-overview.md`
- **CLI Reference**（documentation）：Complete reference for all MnemoQ command-line tools. 证据：`docs/cli-reference.md`
- **Config Tuning Guide**（documentation）：config.json provides project-specific tuning overlaid on engine defaults. Every parameter below has a sensible default in src/agent memory/engine/constants.py ; your memory/config.json overrides only what you need. See templates/config.json for the full template and templates/config-presets/ for ready-made presets. 证据：`docs/config-tuning.md`
- **Data Schema Reference**（documentation）：Canonical reference for the learnings.jsonl entry schema. Each line is a JSON object conforming to the fields below. 证据：`docs/data-schema.md`
- **MCP Integration Guide**（documentation）：The Model Context Protocol MCP is the primary integration path for AI agents to access MnemoQ's memory engine. This guide covers installation, client configuration, tool reference, and troubleshooting. 证据：`docs/mcp-integration.md`
- **Open-Core Architecture**（documentation）：MnemoQ uses an open-core model: the AGPL-3.0-or-later core lives in this public repo, while a proprietary Pro tier runs from a separate private repo. 证据：`docs/open-core-architecture.md`
- **Python SDK Guide**（documentation）：Programmatic access to the MnemoQ memory engine — log, retrieve, update, resolve, consolidate, and read metrics from Python code. 证据：`docs/sdk-guide.md`
- **Init**（source_file）：version = get engine version def getattr name all = "MemoryClient", "LearningEntry", " version " 证据：`src/agent_memory/__init__.py`
- **Whitelist of tuning parameters with type and range validation**（source_file）：@dataclass frozen=True class Paths ⋮---- memory dir: str repo root: str config path: str learnings path: str quarantine path: str archive dir: str session file: str agents md path: str ENGINE VERSION = get engine version PATHS: Paths None = None def get paths - Paths ⋮---- CTX = {k.lower : v for k, v in CONST DEFAULTS.items } def resolve memory dir memory dir arg: str None - str ⋮---- raw = memory dir arg.strip ⋮---- env dir = os.environ.get "AGENT MEMORY DIR" ⋮---- raw = env dir.strip ⋮---- cwd memory = os.path.join os.getcwd , "memory" ⋮---- def setup paths memory dir arg: str None - Paths ⋮---- memory dir = resolve memory dir memory dir arg repo root = os.path.dirname memory dir ⋮---- de… 证据：`src/agent_memory/cli.py`
- **Output report**（source_file）：STOP WORDS = { def tokenize keywords text ⋮---- words = text.lower .split keywords = set ⋮---- word = word.strip ".,;:!? {}\"'" ⋮---- def parse agents sections agents md path ⋮---- """Parse AGENTS.md into a list of heading, content tuples. Extracts , , and headings. Content is everything from after the heading until the next heading of equal or higher level. Returns empty list if file doesn't exist or has no headings. Extract a set of lowercase keywords from a section heading and content. Strips code blocks triple-backtick , extracts table cell keywords split on , strips inline backticks preserving content , lowercases, splits on whitespace, and removes stop-words. Check if a learning overl… 证据：`src/agent_memory/engine/agents_review.py`
- **Consolidation**（source_file）：def sprint metrics paths ⋮---- events = read metrics paths ⋮---- retrievals = e for e in events if e.get "event type" == "retrieval" logs = e for e in events if e.get "event type" == "log" lines = ⋮---- hits = sum 1 for e in retrievals avg lat = sum e.get "latency ms", 0 or 0 for e in retrievals / len retrievals ⋮---- dups = sum 1 for e in logs if e.get "outcome" == "DUPLICATE" quars = sum 1 for e in logs if e.get "outcome" == "QUARANTINED" ⋮---- all lat = e.get "latency ms", 0 or 0 for e in events if e.get "latency ms" ⋮---- def score for promotion entry, current step, ctx ⋮---- access count = entry.get "access count", 0 severity = entry.get "severity", "minor" step diff = current step - e… 证据：`src/agent_memory/engine/consolidation.py`
- **Flat dict of all defaults, keyed by the UPPERCASE names used in globals .update**（source_file）：SESSION EXPIRY MINUTES = 10 DECAY RATE = 0.995 SCORE THRESHOLD = 0.15 COMPONENT WEIGHT = 1.0 FILE WEIGHT = 0.7 DOMAIN WEIGHT = 0.4 NO MATCH WEIGHT = 0.1 MAX WARNINGS = 5 MAX PATTERNS = 15 BM25 K1 = 1.5 BM25 B = 0.75 RRF K = 60 MINOR RETENTION = 5 MAJOR RETENTION = 20 ESCALATION THRESHOLD = 30 SLEEP CYCLE DAYS = 7 SLEEP CYCLE QUARANTINE THRESHOLD = 20 EMBEDDING MODEL = "all-MiniLM-L6-v2" EMBEDDING ALPHA = 0.5 EMBEDDING CACHE DIR = "~/.agent-memory/models/" SEMANTIC DEDUP THRESHOLD = 0.85 RERANKER = "none" RERANKER TOP N = 20 RERANKER MODEL = "cross-encoder/ms-marco-MiniLM-L-12-v2" RERANKER LLM ENDPOINT = None RERANKER LLM MODEL = None API KEY = None VALID RERANKERS = {"none", "cross-encoder"… 证据：`src/agent_memory/engine/constants.py`
- **ponytail: invoke the installed cli module directly; replaces old src/filter.py path math**（source_file）：def load fixture fixture path ⋮---- fixtures = ⋮---- line = line.strip ⋮---- def run eval paths, ctx ⋮---- """Run the grading harness. Reads memory/eval/grading.jsonl, runs retrieval for each fixture, and prints a hit-rate report. ctx is accepted for API consistency with other engine functions but is not used directly — each fixture spawns a subprocess that loads its own config from the memory directory. Returns 0 on success, 1 if no fixtures found. """ fixture path = os.path.join paths.memory dir, "eval", "grading.jsonl" fixtures = load fixture fixture path ⋮---- ponytail: invoke the installed cli module directly; replaces old src/filter.py path math python exe = sys.executable total = len… 证据：`src/agent_memory/engine/eval.py`
- **Parse diff stat output to count lines changed**（source_file）：def stamp entry entry, repo root ⋮---- commit = subprocess.check output ⋮---- commit = "unknown" ⋮---- def check staleness entry, repo root ⋮---- commit = entry "commit" files = entry "files touched" ⋮---- result = subprocess.run ⋮---- Parse diff stat output to count lines changed lines changed = 0 ⋮---- parts = line.split " " ⋮---- changes str = parts 1 .strip .split 0 ⋮---- is stale = lines changed 500 证据：`src/agent_memory/engine/git_utils.py`
- **Re-compute embedding if trigger/action/reason changed**（source_file）：TS PATTERN = re.compile r"^\d{4}-\d{2}-\d{2}T\d{2}:\d{2}:\d{2}Z$" def log core json str, paths, ctx ⋮---- """Shared logic for --log mode. Returns dict, no printing. Returns: {"exit code": int, "status": str, "message": str, ...} Handle --log mode: validate, dedup-check, append. CLI wrapper.""" result = log core json str, paths, ctx ⋮---- def update core ts, json str, paths, ctx ⋮---- """Shared logic for --update mode. Returns dict, no printing.""" start = time.perf counter ⋮---- entry = json.loads json str ⋮---- errors = validate entry entry, ctx ⋮---- reason = "; ".join errors ⋮---- valid retrieval only = ctx.get "valid retrieval only agents" ⋮---- detail = f"{entry 'source agent' } is ret… 证据：`src/agent_memory/engine/handlers.py`
- **---------------------------------------------------------------------------**（source_file）：@dataclass frozen=True class Paths ⋮---- memory dir: str repo root: str config path: str learnings path: str quarantine path: str archive dir: str session file: str agents md path: str def resolve memory dir memory dir arg: str None - str ⋮---- raw = memory dir arg.strip ⋮---- env dir = os.environ.get "AGENT MEMORY DIR" ⋮---- raw = env dir.strip ⋮---- cwd memory = os.path.join os.getcwd , "memory" ⋮---- def setup paths memory dir arg: str None - Paths ⋮---- memory dir = resolve memory dir memory dir arg repo root = os.path.dirname memory dir ⋮---- def load config paths: Paths - dict ⋮---- config path = Path paths.config path ⋮---- config = json.load f ⋮---- result = {} tuning = config.get "… 证据：`src/agent_memory/engine/mcp_server.py`
- **Metrics**（source_file）：ENGINE DIR = Path.home / ".agent-memory" / "engine" ProjectPaths = namedtuple " ProjectPaths", def metrics path paths def get project id paths ⋮---- config path = Path paths.config path ⋮---- config = json.load f name = config.get "project name" ⋮---- def log event paths, event type, fields ⋮---- event = { ⋮---- def read metrics paths, event type=None, since=None ⋮---- path = metrics path paths ⋮---- events = ⋮---- line = line.strip ⋮---- event = json.loads line ⋮---- ts = datetime.fromisoformat event "ts" .replace "Z", "+00:00" ⋮---- def load project paths ⋮---- projects file = ENGINE DIR / "projects.txt" ⋮---- projects = ⋮---- s = line.strip ⋮---- p = Path s ⋮---- def read all project met… 证据：`src/agent_memory/engine/metrics.py`
- **Models**（source_file）：class LearningEntry BaseModel ⋮---- step: int = Field ge=1 source agent: str type: str domain: str components: list str = Field min length=1 files touched: list str = Field min length=1 trigger: str action: str reason: str importance: int = Field ge=1, le=10 severity: str scope: str None = None verified: bool None = None symptoms: str None = None debt level: str None = None schema version: int None = None access count: int None = None reinforcement count: int None = None commit: str None = None embedding: str None = None project id: str None = None origin project: str None = None contributing projects: list str None = None contributors: list str None = None model config = {"extra": "allow"}… 证据：`src/agent_memory/engine/models.py`
- **--- Embedding functions ---**（source_file）：TOKEN SPLIT = re.compile r' ^a-z0-9 +' --- Embedding functions --- embedder cache: dict str, object None = {} def get embedder model name=EMBEDDING MODEL, cache dir=EMBEDDING CACHE DIR ⋮---- """Lazy singleton for sentence-transformers model, keyed by model name. Returns model instance or None cached per model name so failed loads don't retry every call . If model name changes between calls, re-initializes. Compute embedding vector for text. Returns list float or None if model unavailable.""" model = get embedder model name, cache dir ⋮---- vec = model.encode text, convert to numpy=True ⋮---- def cosine similarity vec a, vec b ⋮---- """Stdlib cosine similarity. No numpy required for small ve… 证据：`src/agent_memory/engine/retrieval.py`
- **Engine Version**（source_file）：FALLBACK VERSION = "1.20.6" def read version file - str None ⋮---- v = candidate.read text .strip ⋮---- def get engine version - str ⋮---- v = pkg version "mnemoq" ⋮---- v = read version file 证据：`src/agent_memory/engine_version.py`
- **Mcp Main**（source_file）：def main ⋮---- memory dir = None args = sys.argv 1: ⋮---- idx = args.index "--memory-dir" ⋮---- memory dir = args idx + 1 证据：`src/agent_memory/mcp_main.py`
- **Check if cwd looks like a project**（source_file）：ENGINE VERSION = get engine version ENGINE DIR = Path.home / ".agent-memory" / "engine" def check prerequisites ⋮---- required files = "filter.py", "templates/config.json" ⋮---- path = ENGINE DIR / f ⋮---- def resolve target path cli path ⋮---- """Resolve target project path from CLI arg or cwd.""" ⋮---- target = Path cli path .resolve ⋮---- target = Path.cwd Check if cwd looks like a project ⋮---- def copy engine files target memory, force ⋮---- """Write shim to target project's memory directory.""" ⋮---- Remove profile.py if it exists no longer needed profile path = target memory / "profile.py" ⋮---- shim path = target memory / "filter.py" ⋮---- def prompt project name default name ⋮----… 证据：`src/agent_memory/scaffold.py`
- **Init**（source_file）：all = 证据：`src/agent_memory/sdk/__init__.py`
- **Client**（source_file）：def build ctx paths ⋮---- config = filter.load config ctx = {k.lower : v for k, v in DEFAULTS.items } ⋮---- def validate entry entry: dict str, Any - dict str, Any ⋮---- validated = LearningEntry entry ⋮---- def local error from result result: dict str, Any , ref: str None = None - None ⋮---- status = result.get "status", "error" message = result.get "message", f"engine error: {status}" kwargs = { ⋮---- class LocalTransport ⋮---- def init self, paths, ctx def retrieve self, step, components, files, domain def log self, entry ⋮---- result = log core json.dumps entry , self.paths, self.ctx ⋮---- def update self, ts, entry ⋮---- result = update core ts, json.dumps entry , self.paths, self.ctx… 证据：`src/agent_memory/sdk/client.py`
- **Exceptions**（source_file）：class MemoryError Exception class ValidationError MemoryError ⋮---- """Raised when an entry or request fails validation.""" class NotFoundError MemoryError ⋮---- """Raised when a referenced learning entry does not exist.""" class ConflictError MemoryError ⋮---- """Raised when an operation conflicts with existing state.""" class APIError MemoryError ⋮---- """Raised for unexpected HTTP or internal server errors.""" 证据：`src/agent_memory/sdk/exceptions.py`
- **Contributor License Agreement**（documentation）：Thank you for your interest in contributing to MnemoQ "the Project" . 证据：`CLA.md`
- **Security Policy**（documentation）：Do not open a public issue for security vulnerabilities. 证据：`SECURITY.md`
- **Handoff - 2026-06-25**（documentation）：Session Summary Update this section at end of each session 证据：`memory/HANDOFF.md`
- **System Invariants**（documentation）：Consolidated structural rules. IMMUTABLE during active tasks. Only updated during Sleep Cycle. 证据：`memory/SYSTEM_INVARIANTS.md`
- **Memory**（documentation）：Session start: memory/HANDOFF.md and memory/SYSTEM INVARIANTS.md are auto-loaded by your IDE/agent platform. Act on HANDOFF's "next action" line if present. 证据：`templates/agents-memory-section.md`
- **Your Mission**（documentation）：You are the Fuzzer. A feature has just been implemented. Your job is to try and break it. 证据：`.opencode/Prompts/Fuzzer.md`
- **Review Priority Order**（documentation）：You are a senior code reviewer for this project. You review diffs against this project's engineering rules and produce a structured report with severity-ranked findings. 证据：`.opencode/Prompts/code-reviewer.md`
- **Role and Identity**（documentation）：Role and Identity You are GM , the primary co-developer and orchestrator for this project. You are highly autonomous and strictly action-oriented. You exist to build exceptional software alongside the human developer, keeping their session entirely clean of tool noise. 证据：`.opencode/Prompts/gm.md`
- **Data Sources read all before analysis**（documentation）：You are the Meta-agent. Your sole purpose is to analyze agent performance data and evolve agent prompts to eliminate recurring failures. 证据：`.opencode/Prompts/meta-agent.md`
- **Advanced Metrics & Analytics System**（documentation）：Advanced Metrics & Analytics System 证据：`.windsurf/Plans/advanced-metrics-system-686f76.md`
- **Auto-Learning System**（documentation）：Plan ID: cb6d42 — saved in-repo for consistency with other plans. Audit pass 1d2cfd applied: reworked the retrieval filter data-loss fix , check staleness ctx guard, consolidate gating, API cache/alert hooks, and accuracy notes. See the audit report for rationale. 证据：`.windsurf/Plans/auto-learning-system-cb6d42.md`
- **Fake Memory Generator**（documentation）：A script to bulk-generate valid synthetic memory entries for stress-testing the Agent Memory Engine pipeline, with direct-file and full-pipeline modes. 证据：`.windsurf/Plans/fake-memory-generator-8e535e.md`
- **Memory Engine GUI — Web Dashboard, Tauri Desktop, TUI**（documentation）：Memory Engine GUI — Web Dashboard, Tauri Desktop, TUI 证据：`.windsurf/Plans/memory-engine-gui-686f76.md`
- **Agent Memory Engine — Consolidated Roadmap**（documentation）：Agent Memory Engine — Consolidated Roadmap 证据：`.windsurf/Plans/memory-engine-roadmap-consolidated.md`
- **Memory Upgrade Progression — Sequential Implementation Guide**（documentation）：Memory Upgrade Progression — Sequential Implementation Guide 证据：`.windsurf/Plans/memory-upgrade-progression.md`
- **Tier 1 — Free Tier Quality Foundation v1.17 – v1.19**（documentation）：Tier 1 — Free Tier Quality Foundation v1.17 – v1.19 证据：`.windsurf/Plans/tier-1-expanded.md`
- **Tier 2 — Distribution & Access v1.20 – v1.21**（documentation）：Tier 2 — Distribution & Access v1.20 – v1.21 证据：`.windsurf/Plans/tier-2-expanded.md`
- **Web Dashboard 2.3 — Audit Follow-Up Plan**（documentation）：Web Dashboard 2.3 — Audit Follow-Up Plan 证据：`.windsurf/Plans/web-dashboard-2.3-audit-follow-up.md`
- **Web Dashboard 2.3 — Audit Report**（documentation）：Date: 2026-06-24 Scope: Dashboard backend src/engine/server.py , dashboard api.py , analysis.py , metrics.py and frontend src/dashboard/static/ Test run: python -m pytest tests/ -q → 125 passed 证据：`.windsurf/Plans/web-dashboard-2.3-audit-report.md`
- **Web Dashboard 2.3 — Three-Phase Implementation Plan**（documentation）：Web Dashboard 2.3 — Three-Phase Implementation Plan 证据：`.windsurf/Plans/web-dashboard-2.3.md`
- **Review Priority Order**（documentation）：You are a senior code reviewer for this project. You review diffs against this project's engineering rules and produce a structured report with severity-ranked findings. 证据：`.windsurf/workflows/code-reviewer.md`
- **Your Mission**（documentation）：You are the Docs Writer for AgentMemoryEngine. Your job is to keep READMEs, API docs, and inline comments in sync with code changes. 证据：`.windsurf/workflows/docs-writer.md`
- **Your Mission**（documentation）：You are the Explorer for AgentMemoryEngine. Your job is to map how a feature works across the codebase and return a focused, structured summary. You are cheap, high-volume reading — the parent agent plans against your summary. 证据：`.windsurf/workflows/explorer.md`
- **Your Mission**（documentation）：You are the Fuzzer. A feature has just been implemented for AgentMemoryEngine. Your job is to try and break it. 证据：`.windsurf/workflows/fuzzer.md`
- **Core Directives**（documentation）：You are GM , the primary co-developer and orchestrator for this project. You are highly autonomous and strictly action-oriented. You exist to build exceptional software alongside the human developer, keeping their session entirely clean of tool noise. 证据：`.windsurf/workflows/gm.md`
- **Environment**（documentation）：You are the Meta-agent. Your sole purpose is to analyze agent performance data and evolve agent prompts to eliminate recurring failures for AgentMemoryEngine. 证据：`.windsurf/workflows/meta-agent.md`
- 其余 17 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/README.md`, `README.md`, `CONTRIBUTING.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/README.md`, `README.md`, `CONTRIBUTING.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。

- **Project Overview & System Architecture**：importance `high`
  - source_paths: README.md, docs/README.md, docs/architecture-overview.md, docs/open-core-architecture.md, src/agent_memory/__init__.py
- **Core Engine: Retrieval, Validation, and Consolidation**：importance `high`
  - source_paths: src/agent_memory/cli.py, src/agent_memory/engine/handlers.py, src/agent_memory/engine/retrieval.py, src/agent_memory/engine/validation.py, src/agent_memory/engine/consolidation.py
- **MCP Server, SDK, and Platform Integration**：importance `high`
  - source_paths: src/agent_memory/engine/mcp_server.py, src/agent_memory/mcp_main.py, src/agent_memory/sdk/client.py, src/agent_memory/sdk/__init__.py, src/agent_memory/sdk/exceptions.py
- **Data Schema, Configuration & Deployment**：importance `medium`
  - source_paths: memory/config.json, src/agent_memory/engine/constants.py, src/agent_memory/engine/models.py, src/agent_memory/engine/migrate.py, src/agent_memory/engine/eval.py

## Repo Inspection Evidence / 源码检查证据

- repo_clone_verified: true
- repo_inspection_verified: true
- repo_commit: `ffc1fdc2897372987a25bacec8d44a16fc229957`
- inspected_files: `README.md`, `pyproject.toml`, `docs/README.md`, `docs/ROADMAP.md`, `docs/architecture-overview.md`, `docs/cli-reference.md`, `docs/config-tuning.md`, `docs/data-schema.md`, `docs/mcp-integration.md`, `docs/open-core-architecture.md`, `docs/sdk-guide.md`, `src/agent_memory/__init__.py`, `src/agent_memory/cli.py`, `src/agent_memory/dashboard/__init__.py`, `src/agent_memory/dashboard/static/js/api.js`, `src/agent_memory/dashboard/static/js/app.js`, `src/agent_memory/dashboard/static/js/consolidation.js`, `src/agent_memory/dashboard/static/js/dashboard.js`, `src/agent_memory/dashboard/static/js/events.js`, `src/agent_memory/dashboard/static/js/fleet.js`

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: 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/Mnemoq/MnemoQ | host_targets=mcp_host, claude_code, claude, 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 2: 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/Mnemoq/MnemoQ | 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 3: 运行可能依赖外部服务

- Trigger: 项目说明出现 external service/cloud/webhook/database 等运行依赖关键词。
- Host AI rule: 确认是否有离线 demo、mock 数据或可替代服务。
- Why it matters: 本地安装成功不等于能力可用，外部服务不可用会阻断体验。
- Evidence: packet_text.keyword_scan | https://github.com/Mnemoq/MnemoQ | matched external service / cloud / webhook / database keyword
- 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/Mnemoq/MnemoQ | 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/Mnemoq/MnemoQ | 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/Mnemoq/MnemoQ | 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: 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 | https://github.com/Mnemoq/MnemoQ | 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 8: 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 | https://github.com/Mnemoq/MnemoQ | 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。