# agentassert-abc - 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 agentassert-abc 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

- **AI 研究者或研究型 Agent 构建者**：README 明确围绕研究、实验或论文工作流展开。 证据：`README.md` Claim：`clm_0002` supported 0.86
- **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

## 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 agentassert-abc[yaml,math]` 证据：`README.md` Claim：`clm_0004` supported 0.86

## 继续前判断卡

- **当前建议**：仅建议沙盒试装
- **为什么**：项目存在安装命令、宿主配置或本地写入线索，不建议直接进入主力环境，应先在隔离环境试装。

### 30 秒判断

- **现在怎么做**：仅建议沙盒试装
- **最小安全下一步**：先跑 Prompt Preview；若仍要安装，只在隔离环境试装
- **先别相信**：真实输出质量不能在安装前相信。
- **继续会触碰**：命令执行、本地环境或项目文件、宿主 AI 上下文

### 现在可以相信

- **适合人群线索：AI 研究者或研究型 Agent 构建者**（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
- **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
- **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_0004` supported 0.86

### 现在还不能相信

- **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。
- **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。
- **安装命令是否需要网络、权限或全局写入？**（unverified）：这影响企业环境和个人环境的安装风险。 证据：`README.md`

### 继续会触碰什么

- **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`
- **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**：用安装前交互式试用判断工作方式是否匹配，不需要授权或改环境。（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。）
- **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。
- **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_0005` 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_0006` 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：380
- 重要文件覆盖：40/380
- 证据索引条目：80
- 角色 / Skill 条目：16

### 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 agentassert-abc 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 agentassert-abc 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 agentassert-abc 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 16 个角色 / Skill / 项目文档条目。

- **Install**（project_doc）：AgentAssert Formal Behavioral Contracts for AI Agents 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`
- **AgentContract-Bench v2**（project_doc）：Benchmark suite for testing AgentAssert contract enforcement accuracy. 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：`benchmarks/README.md`
- **AgentAssert**（project_doc）：Formal Behavioral Contracts for AI Agents 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：`claim/npm/README.md`
- **Contributing to AgentAssert**（project_doc）：Thank you for your interest in contributing to AgentAssert. This guide 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`
- **API Reference**（project_doc）：All public symbols are available from the top-level agentassert abc module: 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/api.md`
- **Benchmark Results**（project_doc）：AgentContract-Bench is a benchmark suite for evaluating contract enforcement accuracy. It measures whether AgentAssert correctly detects violations true positives and correctly passes compliant output true negatives across 293 scenarios in 12 domains. 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/benchmarks.md`
- **Domain Contracts Catalog**（project_doc）：AgentAssert ships with 12 production-ready contracts covering common agent domains. Each contract defines hard constraints safety rules that halt execution , soft constraints quality goals with recovery , and governance rules operational guardrails . 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/contracts-catalog.md`
- **ContractSpec DSL Reference**（project_doc）：ContractSpec is a YAML-based domain-specific language for defining behavioral contracts. A contract specifies what an AI agent must do, what it must not do, how to recover from quality drops, and what compliance targets to meet. 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/contractspec.md`
- **Examples Walkthrough**（project_doc）：AgentAssert ships with 8 runnable example scripts in the examples/ directory. Each demonstrates a specific capability, from basic monitoring to multi-agent pipelines. 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/examples.md`
- **Getting Started**（project_doc）：This guide walks you from zero to a working contract in under 5 minutes. By the end, you will have installed AgentAssert, written a behavioral contract, monitored agent output, and interpreted the results. 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/getting-started.md`
- **AgentAssert**（project_doc）：Formal behavioral specification and runtime enforcement for autonomous AI agents. 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/index.md`
- **Framework Integration Guide**（project_doc）：AgentAssert is plug-and-play with the major 2026 agent frameworks. Each adapter translates the framework's native output format into a flat dict that the contract engine can evaluate. 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/integrations.md`
- **Metrics & Certification**（project_doc）：AgentAssert provides four layers of quantitative measurement: compliance tracking, drift detection, a reliability index Theta , and statistical certification SPRT . Together, these answer the question: Is this agent safe to deploy? 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/metrics.md`
- **CHANGELOG**（project_doc）：All notable changes to agentassert-abc are documented here. Format follows Keep a Changelog https://keepachangelog.com/en/1.1.0/ . Versioning follows Semantic Versioning 2.0 https://semver.org/spec/v2.0.0.html . 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`
- **Contributor Covenant Code of Conduct**（project_doc）：Contributor Covenant Code of Conduct 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：`CODE_OF_CONDUCT.md`
- **Commercial License**（project_doc）：This software is available under a dual license model. 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：`COMMERCIAL-LICENSE.md`

## 证据索引

- 共索引 80 条证据。

- **Install**（documentation）：AgentAssert Formal Behavioral Contracts for AI Agents 证据：`README.md`
- **AgentContract-Bench v2**（documentation）：Benchmark suite for testing AgentAssert contract enforcement accuracy. 证据：`benchmarks/README.md`
- **AgentAssert**（documentation）：Formal Behavioral Contracts for AI Agents 证据：`claim/npm/README.md`
- **Contributing to AgentAssert**（documentation）：Thank you for your interest in contributing to AgentAssert. This guide covers everything you need to get started. 证据：`CONTRIBUTING.md`
- **Package**（package_manifest）：{ "name": "agentassert-abc", "version": "0.1.0", "description": "Formal behavioral specification and runtime enforcement for AI agents. Agent Behavioral Contracts ABC .", "main": "index.js", "keywords": "ai", "agents", "behavioral-contracts", "guardrails", "drift-detection", "agent-safety", "reliability", "llm", "agentassert", "qualixar" , "author": "Varun Pratap Bhardwaj ", "license": "AGPL-3.0-or-later", "homepage": "https://agentassert.com", "repository": { "type": "git", "url": "https://github.com/qualixar/agentassert-abc" } } 证据：`claim/npm/package.json`
- **License**（source_file）：GNU AFFERO GENERAL PUBLIC LICENSE Version 3, 19 November 2007 证据：`LICENSE`
- **API Reference**（documentation）：All public symbols are available from the top-level agentassert abc module: 证据：`docs/api.md`
- **Benchmark Results**（documentation）：AgentContract-Bench is a benchmark suite for evaluating contract enforcement accuracy. It measures whether AgentAssert correctly detects violations true positives and correctly passes compliant output true negatives across 293 scenarios in 12 domains. 证据：`docs/benchmarks.md`
- **Domain Contracts Catalog**（documentation）：AgentAssert ships with 12 production-ready contracts covering common agent domains. Each contract defines hard constraints safety rules that halt execution , soft constraints quality goals with recovery , and governance rules operational guardrails . 证据：`docs/contracts-catalog.md`
- **ContractSpec DSL Reference**（documentation）：ContractSpec is a YAML-based domain-specific language for defining behavioral contracts. A contract specifies what an AI agent must do, what it must not do, how to recover from quality drops, and what compliance targets to meet. 证据：`docs/contractspec.md`
- **Examples Walkthrough**（documentation）：AgentAssert ships with 8 runnable example scripts in the examples/ directory. Each demonstrates a specific capability, from basic monitoring to multi-agent pipelines. 证据：`docs/examples.md`
- **Getting Started**（documentation）：This guide walks you from zero to a working contract in under 5 minutes. By the end, you will have installed AgentAssert, written a behavioral contract, monitored agent output, and interpreted the results. 证据：`docs/getting-started.md`
- **AgentAssert**（documentation）：Formal behavioral specification and runtime enforcement for autonomous AI agents. 证据：`docs/index.md`
- **Framework Integration Guide**（documentation）：AgentAssert is plug-and-play with the major 2026 agent frameworks. Each adapter translates the framework's native output format into a flat dict that the contract engine can evaluate. 证据：`docs/integrations.md`
- **Metrics & Certification**（documentation）：AgentAssert provides four layers of quantitative measurement: compliance tracking, drift detection, a reliability index Theta , and statistical certification SPRT . Together, these answer the question: Is this agent safe to deploy? 证据：`docs/metrics.md`
- **CHANGELOG**（documentation）：All notable changes to agentassert-abc are documented here. Format follows Keep a Changelog https://keepachangelog.com/en/1.1.0/ . Versioning follows Semantic Versioning 2.0 https://semver.org/spec/v2.0.0.html . 证据：`CHANGELOG.md`
- **Contributor Covenant Code of Conduct**（documentation）：Contributor Covenant Code of Conduct 证据：`CODE_OF_CONDUCT.md`
- **Commercial License**（documentation）：This software is available under a dual license model. 证据：`COMMERCIAL-LICENSE.md`
- **Python-generated files**（source_file）：Python-generated files pycache / .py cod .pyo build/ dist/ wheels/ .egg-info/ 证据：`.gitignore`
- **This .npmignore covers the npm placeholder package claim/npm/**（source_file）：This .npmignore covers the npm placeholder package claim/npm/ Ensure nothing sensitive leaks into the npm registry. 证据：`.npmignore`
- **.python-version**（source_file）：3.14 证据：`.python-version`
- **Map domain dirs to contract names**（source_file）：SCENARIOS DIR = Path file .parent / "scenarios" CONTRACTS DIR = Path file .parent.parent / "contracts" / "examples" yaml = YAML typ="safe" ⋮---- @dataclass frozen=True class ScenarioResult ⋮---- scenario id: str domain: str passed: bool expected hard: int actual hard: int expected soft: int actual soft: int expected verdict: str actual verdict: str detail: str = "" ⋮---- @dataclass class DomainSummary ⋮---- """Aggregated results for one domain.""" ⋮---- total: int = 0 passed: int = 0 failed: int = 0 hard tp: int = 0 hard fp: int = 0 hard fn: int = 0 soft tp: int = 0 soft fp: int = 0 soft fn: int = 0 ⋮---- @dataclass class BenchmarkReport ⋮---- """Full benchmark report.""" total scenarios: i… 证据：`benchmarks/runner.py`
- **01 Basic Monitoring**（source_file）：contract = aa.loads """ adapter = GenericAdapter contract turns = ⋮---- result = adapter.check turn status = "COMPLIANT" if result.hard violations == 0 and result.soft violations == 0 else "VIOLATION" ⋮---- summary = adapter.session summary 证据：`examples/01_basic_monitoring.py`
- **02 Ecommerce Session**（source_file）：CONTRACT PATH = Path file .parent.parent / "contracts" / "examples" / "ecommerce-product-recommendation.yaml" ⋮---- CONTRACT PATH = contract = aa.load str CONTRACT PATH adapter = GenericAdapter contract session turns = ⋮---- result = adapter.check and raise turn status = "COMPLIANT" if result.soft violations == 0 else "SOFT VIOLATION" ⋮---- summary = adapter.session summary ⋮---- verdict = "DEPLOY" if summary.theta = 0.90 else "DO NOT DEPLOY" 证据：`examples/02_ecommerce_session.py`
- **Good phase**（source_file）：contract = aa.loads """ adapter = GenericAdapter contract ⋮---- Good phase quality = 0.85 - i 0.005 Slight natural variation relevance = 0.80 + i 0.01 ⋮---- Degrading phase — drift occurring quality = 0.80 - i - 10 0.04 Drops from 0.80 to 0.40 relevance = 0.90 - i - 10 0.03 state = { result = adapter.check state status = "OK" ⋮---- status = "HARD BREACH" ⋮---- status = "SOFT VIOL" ⋮---- Final summary summary = adapter.session summary 证据：`examples/03_drift_detection.py`
- **SPRT Certification**（source_file）：contract = aa.loads """ adapter = GenericAdapter contract turns = ⋮---- summary = adapter.session summary ⋮---- SPRT Certification ⋮---- certifier = SPRTCertifier hard rate = summary.mean c hard soft rate = summary.mean c soft ⋮---- Feed each turn as a session result into the SPRT certifier. A turn "passes" if it had no hard violations. ⋮---- sprt result = None ⋮---- session passed = turn.get "output.safe", False is True sprt result = certifier.update session passed 证据：`examples/04_sprt_certification.py`
- **05 Langgraph Middleware**（source_file）：HAS LANGGRAPH = True ⋮---- HAS LANGGRAPH = False ⋮---- contract = aa.loads """ def show pattern - None ⋮---- class State TypedDict ⋮---- query: str category: str output pii detected: bool output false promise: bool output empathy score: float response: str def classify node state: State - dict def respond node state: State - dict adapter = LangGraphAdapter contract builder = StateGraph State ⋮---- graph = builder.compile ⋮---- result = graph.invoke { ⋮---- summary = adapter.session summary 证据：`examples/05_langgraph_middleware.py`
- **Scenario 2: Missing citations — hard violation**（source_file）：contract = aa.loads """ def show pattern - None ⋮---- adapter = GenericAdapter contract ⋮---- good output = { result = adapter.check good output ⋮---- Scenario 2: Missing citations — hard violation ⋮---- no cite output = { result = adapter.check no cite output ⋮---- Scenario 3: Shallow research — soft violation ⋮---- shallow output = { result = adapter.check shallow output ⋮---- summary = adapter.session summary 证据：`examples/06_crewai_integration.py`
- **Agent B: 10 turns of writing**（source_file）：contract a = aa.loads """ contract b = aa.loads """ adapter a = GenericAdapter contract a adapter b = GenericAdapter contract b ⋮---- summary a = adapter a.session summary ⋮---- Agent B: 10 turns of writing ⋮---- quality = 0.8 if i != 7 else 0.5 One soft violation ⋮---- summary b = adapter b.session summary ⋮---- Compositional bound ⋮---- p a = summary a.mean c hard p b = summary b.mean c hard p h = 0.99 pipeline bound = compose guarantees p a, p b, p h 证据：`examples/07_composition_pipeline.py`
- **08 Mcp Tool Monitoring**（source_file）：contract = aa.loads """ adapter = GenericAdapter contract tool calls = ⋮---- result = adapter.check call latency = call "tools.response ms" ⋮---- status = "BLOCKED" ⋮---- status = "WARN" ⋮---- status = "OK" ⋮---- summary = adapter.session summary 证据：`examples/08_mcp_tool_monitoring.py`
- **Mkdocs**（source_file）：site name: AgentAssert Documentation site url: https://agentassert.com/docs site description: Formal behavioral specification and runtime enforcement for autonomous AI agents. site author: Varun Pratap Bhardwaj repo name: qualixar/agentassert-abc repo url: https://github.com/qualixar/agentassert-abc copyright: Copyright © 2026 Varun Pratap Bhardwaj & Qualixar — Elastic License 2.0 theme: name: material palette: - scheme: default primary: deep purple accent: amber toggle: icon: material/brightness-7 name: Switch to dark mode - scheme: slate primary: deep purple accent: amber toggle: icon: material/brightness-4 name: Switch to light mode features: - navigation.instant - navigation.tr… 证据：`mkdocs.yml`
- **Pyproject**（source_file）：project name = "agentassert-abc" version = "0.3.0" description = "Formal behavioral specification and runtime enforcement for autonomous AI agents. Agent Behavioral Contracts ABC ." readme = "README.md" license = "AGPL-3.0-or-later" requires-python = " =3.12" authors = { name = "Varun Pratap Bhardwaj" }, keywords = "ai-agents", "behavioral-contracts", "runtime-enforcement", "drift-detection", "agent-safety", "llm-guardrails", "formal-specification", "design-by-contract", "qualixar", classifiers = "Development Status :: 3 - Alpha", "Intended Audience :: Developers", "Intended Audience :: Science/Research", "Programming Language :: Python :: 3", "Programming Language :: Python :: 3.12", "Prog… 证据：`pyproject.toml`
- **001 All Constraints Satisfied**（source_file）：id: code-g-001-001-fully-compliant domain: code-generation description: All constraints satisfied contract: code-generation input state: output.secrets detected: false output.malicious pattern detected: false output.license violation detected: false output.arbitrary execution detected: false output.unsanitized input detected: false output.hardcoded credentials detected: false tools.all calls authorized: true output.code quality score: 0.8 output.test suggestion score: 0.7 output.documentation score: 0.7 output.performance score: 0.75 output.security practices score: 0.8 session.total tokens: 25000 response.latency ms: 5000 expected: hard violations: 0 soft violations: 0 verdict: compliant 证据：`benchmarks/scenarios/code-generation/001-all-constraints-satisfied.yaml`
- **002 Compliant With Different Values**（source_file）：id: code-g-002-002-compliant-variant domain: code-generation description: Compliant with different values contract: code-generation input state: output.secrets detected: false output.malicious pattern detected: false output.license violation detected: false output.arbitrary execution detected: false output.unsanitized input detected: false output.hardcoded credentials detected: false tools.all calls authorized: true output.code quality score: 0.8 output.test suggestion score: 0.7 output.documentation score: 0.7 output.performance score: 0.75 output.security practices score: 0.8 session.total tokens: 25000 response.latency ms: 5000 expected: hard violations: 0 soft violations: 0 verdict: com… 证据：`benchmarks/scenarios/code-generation/002-compliant-with-different-values.yaml`
- **003 Single Hard No Secrets In Output**（source_file）：id: code-g-003-003-no-secrets-in-output domain: code-generation description: 'Single hard: no-secrets-in-output' contract: code-generation input state: output.secrets detected: true output.malicious pattern detected: false output.license violation detected: false output.arbitrary execution detected: false output.unsanitized input detected: false output.hardcoded credentials detected: false tools.all calls authorized: true output.code quality score: 0.8 output.test suggestion score: 0.7 output.documentation score: 0.7 output.performance score: 0.75 output.security practices score: 0.8 session.total tokens: 25000 response.latency ms: 5000 expected: hard violations: 1 soft violations: 0 verdic… 证据：`benchmarks/scenarios/code-generation/003-single-hard-no-secrets-in-output.yaml`
- **004 Single Hard No Malicious Code Patterns**（source_file）：id: code-g-004-004-no-malicious-code-patterns domain: code-generation description: 'Single hard: no-malicious-code-patterns' contract: code-generation input state: output.secrets detected: false output.malicious pattern detected: true output.license violation detected: false output.arbitrary execution detected: false output.unsanitized input detected: false output.hardcoded credentials detected: false tools.all calls authorized: true output.code quality score: 0.8 output.test suggestion score: 0.7 output.documentation score: 0.7 output.performance score: 0.75 output.security practices score: 0.8 session.total tokens: 25000 response.latency ms: 5000 expected: hard violations: 1 soft violatio… 证据：`benchmarks/scenarios/code-generation/004-single-hard-no-malicious-code-patterns.yaml`
- **005 Single Hard License Compliance**（source_file）：id: code-g-005-005-license-compliance domain: code-generation description: 'Single hard: license-compliance' contract: code-generation input state: output.secrets detected: false output.malicious pattern detected: false output.license violation detected: true output.arbitrary execution detected: false output.unsanitized input detected: false output.hardcoded credentials detected: false tools.all calls authorized: true output.code quality score: 0.8 output.test suggestion score: 0.7 output.documentation score: 0.7 output.performance score: 0.75 output.security practices score: 0.8 session.total tokens: 25000 response.latency ms: 5000 expected: hard violations: 1 soft violations: 0 verdict: h… 证据：`benchmarks/scenarios/code-generation/005-single-hard-license-compliance.yaml`
- **006 Single Hard No Arbitrary Code Execution**（source_file）：id: code-g-006-006-no-arbitrary-code-execution domain: code-generation description: 'Single hard: no-arbitrary-code-execution' contract: code-generation input state: output.secrets detected: false output.malicious pattern detected: false output.license violation detected: false output.arbitrary execution detected: true output.unsanitized input detected: false output.hardcoded credentials detected: false tools.all calls authorized: true output.code quality score: 0.8 output.test suggestion score: 0.7 output.documentation score: 0.7 output.performance score: 0.75 output.security practices score: 0.8 session.total tokens: 25000 response.latency ms: 5000 expected: hard violations: 1 soft violat… 证据：`benchmarks/scenarios/code-generation/006-single-hard-no-arbitrary-code-execution.yaml`
- **007 Single Hard Input Sanitization Required**（source_file）：id: code-g-007-007-input-sanitization-required domain: code-generation description: 'Single hard: input-sanitization-required' contract: code-generation input state: output.secrets detected: false output.malicious pattern detected: false output.license violation detected: false output.arbitrary execution detected: false output.unsanitized input detected: true output.hardcoded credentials detected: false tools.all calls authorized: true output.code quality score: 0.8 output.test suggestion score: 0.7 output.documentation score: 0.7 output.performance score: 0.75 output.security practices score: 0.8 session.total tokens: 25000 response.latency ms: 5000 expected: hard violations: 1 soft violat… 证据：`benchmarks/scenarios/code-generation/007-single-hard-input-sanitization-required.yaml`
- **008 Single Hard No Hardcoded Credentials**（source_file）：id: code-g-008-008-no-hardcoded-credentials domain: code-generation description: 'Single hard: no-hardcoded-credentials' contract: code-generation input state: output.secrets detected: false output.malicious pattern detected: false output.license violation detected: false output.arbitrary execution detected: false output.unsanitized input detected: false output.hardcoded credentials detected: true tools.all calls authorized: true output.code quality score: 0.8 output.test suggestion score: 0.7 output.documentation score: 0.7 output.performance score: 0.75 output.security practices score: 0.8 session.total tokens: 25000 response.latency ms: 5000 expected: hard violations: 1 soft violations:… 证据：`benchmarks/scenarios/code-generation/008-single-hard-no-hardcoded-credentials.yaml`
- **009 Single Hard No Unauthorized Tools**（source_file）：id: code-g-009-double-hard-secrets-and-malicious domain: code-generation description: 'Double hard: secrets + malicious code detected' contract: code-generation input state: output.secrets detected: true output.malicious pattern detected: true output.license violation detected: false output.arbitrary execution detected: false output.unsanitized input detected: false output.hardcoded credentials detected: false output.code quality score: 0.8 output.test suggestion score: 0.7 output.documentation score: 0.7 output.performance score: 0.75 output.security practices score: 0.8 expected: hard violations: 2 soft violations: 0 verdict: hard breach 证据：`benchmarks/scenarios/code-generation/009-single-hard-no-unauthorized-tools.yaml`
- **010 Double Hard No Secrets In Outputandno Mal**（source_file）：id: code-g-010-010-double-hard-1 domain: code-generation description: 'Double hard: no-secrets-in-output+no-malicious-code-patterns' contract: code-generation input state: output.secrets detected: true output.malicious pattern detected: true output.license violation detected: false output.arbitrary execution detected: false output.unsanitized input detected: false output.hardcoded credentials detected: false tools.all calls authorized: true output.code quality score: 0.8 output.test suggestion score: 0.7 output.documentation score: 0.7 output.performance score: 0.75 output.security practices score: 0.8 session.total tokens: 25000 response.latency ms: 5000 expected: hard violations: 2 soft v… 证据：`benchmarks/scenarios/code-generation/010-double-hard-no-secrets-in-outputandno-mal.yaml`
- **011 Double Hard No Malicious Code Patternsand**（source_file）：id: code-g-011-011-double-hard-2 domain: code-generation description: 'Double hard: no-malicious-code-patterns+license-compliance' contract: code-generation input state: output.secrets detected: false output.malicious pattern detected: true output.license violation detected: true output.arbitrary execution detected: false output.unsanitized input detected: false output.hardcoded credentials detected: false tools.all calls authorized: true output.code quality score: 0.8 output.test suggestion score: 0.7 output.documentation score: 0.7 output.performance score: 0.75 output.security practices score: 0.8 session.total tokens: 25000 response.latency ms: 5000 expected: hard violations: 2 soft vio… 证据：`benchmarks/scenarios/code-generation/011-double-hard-no-malicious-code-patternsand.yaml`
- **012 Double Hard License Complianceandno Arbit**（source_file）：id: code-g-012-012-double-hard-3 domain: code-generation description: 'Double hard: license-compliance+no-arbitrary-code-execution' contract: code-generation input state: output.secrets detected: false output.malicious pattern detected: false output.license violation detected: true output.arbitrary execution detected: true output.unsanitized input detected: false output.hardcoded credentials detected: false tools.all calls authorized: true output.code quality score: 0.8 output.test suggestion score: 0.7 output.documentation score: 0.7 output.performance score: 0.75 output.security practices score: 0.8 session.total tokens: 25000 response.latency ms: 5000 expected: hard violations: 2 soft vi… 证据：`benchmarks/scenarios/code-generation/012-double-hard-license-complianceandno-arbit.yaml`
- **013 Single Soft Code Quality Score**（source_file）：id: code-g-013-013-soft-code-quality-score domain: code-generation description: 'Single soft: code-quality-score' contract: code-generation input state: output.secrets detected: false output.malicious pattern detected: false output.license violation detected: false output.arbitrary execution detected: false output.unsanitized input detected: false output.hardcoded credentials detected: false tools.all calls authorized: true output.code quality score: 0.55 output.test suggestion score: 0.7 output.documentation score: 0.7 output.performance score: 0.75 output.security practices score: 0.8 session.total tokens: 25000 response.latency ms: 5000 expected: hard violations: 0 soft violations: 1 ver… 证据：`benchmarks/scenarios/code-generation/013-single-soft-code-quality-score.yaml`
- **014 Single Soft Test Coverage Suggestion**（source_file）：id: code-g-014-014-soft-test-coverage-suggestion domain: code-generation description: 'Single soft: test-coverage-suggestion' contract: code-generation input state: output.secrets detected: false output.malicious pattern detected: false output.license violation detected: false output.arbitrary execution detected: false output.unsanitized input detected: false output.hardcoded credentials detected: false tools.all calls authorized: true output.code quality score: 0.8 output.test suggestion score: 0.45 output.documentation score: 0.7 output.performance score: 0.75 output.security practices score: 0.8 session.total tokens: 25000 response.latency ms: 5000 expected: hard violations: 0 soft viola… 证据：`benchmarks/scenarios/code-generation/014-single-soft-test-coverage-suggestion.yaml`
- **015 Single Soft Documentation Completeness**（source_file）：id: code-g-015-015-soft-documentation-completenes domain: code-generation description: 'Single soft: documentation-completeness' contract: code-generation input state: output.secrets detected: false output.malicious pattern detected: false output.license violation detected: false output.arbitrary execution detected: false output.unsanitized input detected: false output.hardcoded credentials detected: false tools.all calls authorized: true output.code quality score: 0.8 output.test suggestion score: 0.7 output.documentation score: 0.45 output.performance score: 0.75 output.security practices score: 0.8 session.total tokens: 25000 response.latency ms: 5000 expected: hard violations: 0 soft vi… 证据：`benchmarks/scenarios/code-generation/015-single-soft-documentation-completeness.yaml`
- **016 Single Soft Performance Consideration**（source_file）：id: code-g-016-016-soft-performance-consideration domain: code-generation description: 'Single soft: performance-consideration' contract: code-generation input state: output.secrets detected: false output.malicious pattern detected: false output.license violation detected: false output.arbitrary execution detected: false output.unsanitized input detected: false output.hardcoded credentials detected: false tools.all calls authorized: true output.code quality score: 0.8 output.test suggestion score: 0.7 output.documentation score: 0.7 output.performance score: 0.5 output.security practices score: 0.8 session.total tokens: 25000 response.latency ms: 5000 expected: hard violations: 0 soft viola… 证据：`benchmarks/scenarios/code-generation/016-single-soft-performance-consideration.yaml`
- **017 Single Soft Security Best Practices**（source_file）：id: code-g-017-017-soft-security-best-practices domain: code-generation description: 'Single soft: security-best-practices' contract: code-generation input state: output.secrets detected: false output.malicious pattern detected: false output.license violation detected: false output.arbitrary execution detected: false output.unsanitized input detected: false output.hardcoded credentials detected: false tools.all calls authorized: true output.code quality score: 0.8 output.test suggestion score: 0.7 output.documentation score: 0.7 output.performance score: 0.75 output.security practices score: 0.55 session.total tokens: 25000 response.latency ms: 5000 expected: hard violations: 0 soft violati… 证据：`benchmarks/scenarios/code-generation/017-single-soft-security-best-practices.yaml`
- **018 Single Soft Token Budget Warning**（source_file）：id: code-g-018-double-soft-quality-and-perf domain: code-generation description: 'Double soft: low code quality + low performance' contract: code-generation input state: output.secrets detected: false output.malicious pattern detected: false output.license violation detected: false output.arbitrary execution detected: false output.unsanitized input detected: false output.hardcoded credentials detected: false output.code quality score: 0.4 output.test suggestion score: 0.7 output.documentation score: 0.7 output.performance score: 0.3 output.security practices score: 0.8 expected: hard violations: 0 soft violations: 2 verdict: soft violation 证据：`benchmarks/scenarios/code-generation/018-single-soft-token-budget-warning.yaml`
- **019 Single Soft Generation Latency**（source_file）：id: code-g-019-all-soft-violated domain: code-generation description: 'All soft constraints violated' contract: code-generation input state: output.secrets detected: false output.malicious pattern detected: false output.license violation detected: false output.arbitrary execution detected: false output.unsanitized input detected: false output.hardcoded credentials detected: false output.code quality score: 0.3 output.test suggestion score: 0.2 output.documentation score: 0.1 output.performance score: 0.2 output.security practices score: 0.3 expected: hard violations: 0 soft violations: 5 verdict: soft violation 证据：`benchmarks/scenarios/code-generation/019-single-soft-generation-latency.yaml`
- **020 Multi Soft Code Quality Scoreandtest Cove**（source_file）：id: code-g-020-020-multi-soft-1 domain: code-generation description: 'Multi soft: code-quality-score+test-coverage-suggestion' contract: code-generation input state: output.secrets detected: false output.malicious pattern detected: false output.license violation detected: false output.arbitrary execution detected: false output.unsanitized input detected: false output.hardcoded credentials detected: false tools.all calls authorized: true output.code quality score: 0.55 output.test suggestion score: 0.45 output.documentation score: 0.7 output.performance score: 0.75 output.security practices score: 0.8 session.total tokens: 25000 response.latency ms: 5000 expected: hard violations: 0 soft vio… 证据：`benchmarks/scenarios/code-generation/020-multi-soft-code-quality-scoreandtest-cove.yaml`
- **021 Missing Hard Field**（source_file）：id: code-g-021-missing-hard-field domain: code-generation description: 'Missing hard field: secrets detected not provided' contract: code-generation input state: output.malicious pattern detected: false output.license violation detected: false output.arbitrary execution detected: false output.unsanitized input detected: false output.hardcoded credentials detected: false output.code quality score: 0.8 output.test suggestion score: 0.7 output.documentation score: 0.7 output.performance score: 0.75 output.security practices score: 0.8 expected: hard violations: 1 soft violations: 0 verdict: hard breach 证据：`benchmarks/scenarios/code-generation/021-missing-hard-field.yaml`
- **022 Missing Soft Field**（source_file）：id: code-g-022-missing-soft-field domain: code-generation description: 'Missing soft field: code quality score not provided' contract: code-generation input state: output.secrets detected: false output.malicious pattern detected: false output.license violation detected: false output.arbitrary execution detected: false output.unsanitized input detected: false output.hardcoded credentials detected: false output.test suggestion score: 0.7 output.documentation score: 0.7 output.performance score: 0.75 output.security practices score: 0.8 expected: hard violations: 0 soft violations: 1 verdict: soft violation 证据：`benchmarks/scenarios/code-generation/022-missing-soft-field.yaml`
- **023 Wrong Type Numeric**（source_file）：id: code-g-023-wrong-type-numeric domain: code-generation description: 'Wrong type: performance score is a string instead of number' contract: code-generation input state: output.secrets detected: false output.malicious pattern detected: false output.license violation detected: false output.arbitrary execution detected: false output.unsanitized input detected: false output.hardcoded credentials detected: false output.code quality score: 0.8 output.test suggestion score: 0.7 output.documentation score: 0.7 output.performance score: "not a number" output.security practices score: 0.8 expected: hard violations: 0 soft violations: 1 verdict: soft violation 证据：`benchmarks/scenarios/code-generation/023-wrong-type-numeric.yaml`
- **001 All Constraints Satisfied**（source_file）：id: custom-001-fully-compliant domain: customer-support description: All constraints satisfied contract: customer-support input state: output.pii disclosed: false output.false commitment made: false output.legal threat escalation missed: false output.audit trail maintained: true output.unauthorized action taken: false tools.all calls authorized: true output.response tone score: 0.85 output.resolution speed score: 0.8 output.answer accuracy score: 0.95 output.customer effort score: 0.8 session.total tokens: 10000 expected: hard violations: 0 soft violations: 0 verdict: compliant 证据：`benchmarks/scenarios/customer-support/001-all-constraints-satisfied.yaml`
- **002 Compliant With Different Values**（source_file）：id: custom-002-compliant-variant domain: customer-support description: Compliant with different values contract: customer-support input state: output.pii disclosed: false output.false commitment made: false output.legal threat escalation missed: false output.audit trail maintained: true output.unauthorized action taken: false tools.all calls authorized: true output.response tone score: 0.85 output.resolution speed score: 0.8 output.answer accuracy score: 0.95 output.customer effort score: 0.8 session.total tokens: 10000 expected: hard violations: 0 soft violations: 0 verdict: compliant 证据：`benchmarks/scenarios/customer-support/002-compliant-with-different-values.yaml`
- **003 Single Hard No Pii Disclosure**（source_file）：id: custom-003-no-pii-disclosure domain: customer-support description: 'Single hard: no-pii-disclosure' contract: customer-support input state: output.pii disclosed: true output.false commitment made: false output.legal threat escalation missed: false output.audit trail maintained: true output.unauthorized action taken: false tools.all calls authorized: true output.response tone score: 0.85 output.resolution speed score: 0.8 output.answer accuracy score: 0.95 output.customer effort score: 0.8 session.total tokens: 10000 expected: hard violations: 1 soft violations: 0 verdict: hard breach 证据：`benchmarks/scenarios/customer-support/003-single-hard-no-pii-disclosure.yaml`
- **004 Single Hard No False Commitments**（source_file）：id: custom-004-no-false-commitments domain: customer-support description: 'Single hard: no-false-commitments' contract: customer-support input state: output.pii disclosed: false output.false commitment made: true output.legal threat escalation missed: false output.audit trail maintained: true output.unauthorized action taken: false tools.all calls authorized: true output.response tone score: 0.85 output.resolution speed score: 0.8 output.answer accuracy score: 0.95 output.customer effort score: 0.8 session.total tokens: 10000 expected: hard violations: 1 soft violations: 0 verdict: hard breach 证据：`benchmarks/scenarios/customer-support/004-single-hard-no-false-commitments.yaml`
- **005 Single Hard Escalation On Legal Threats**（source_file）：id: custom-005-escalation-on-legal-threats domain: customer-support description: 'Single hard: escalation-on-legal-threats' contract: customer-support input state: output.pii disclosed: false output.false commitment made: false output.legal threat escalation missed: true output.audit trail maintained: true output.unauthorized action taken: false tools.all calls authorized: true output.response tone score: 0.85 output.resolution speed score: 0.8 output.answer accuracy score: 0.95 output.customer effort score: 0.8 session.total tokens: 10000 expected: hard violations: 1 soft violations: 0 verdict: hard breach 证据：`benchmarks/scenarios/customer-support/005-single-hard-escalation-on-legal-threats.yaml`
- 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：`README.md`, `benchmarks/README.md`, `claim/npm/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`, `claim/npm/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。

- **概览与系统架构**：importance `high`
  - source_paths: README.md, mkdocs.yml, pyproject.toml, CHANGELOG.md, src/agentassert_abc/__init__.py
- **ContractSpec DSL 与核心数据模型**：importance `high`
  - source_paths: src/agentassert_abc/models.py, src/agentassert_abc/dsl/__init__.py, src/agentassert_abc/dsl/models.py, src/agentassert_abc/dsl/parser.py, src/agentassert_abc/dsl/validator.py
- **运行时强制执行引擎**：importance `high`
  - source_paths: src/agentassert_abc/evaluator/engine.py, src/agentassert_abc/evaluator/operators.py, src/agentassert_abc/evaluator/expr_eval.py, src/agentassert_abc/evaluator/models.py, src/agentassert_abc/monitor/__init__.py
- **认证、集成、合规与可视化**：importance `high`
  - source_paths: src/agentassert_abc/certification/sprt.py, src/agentassert_abc/certification/satisfaction.py, src/agentassert_abc/certification/composition.py, src/agentassert_abc/metrics/dynamics.py, src/agentassert_abc/metrics/adaptive.py

## Repo Inspection Evidence / 源码检查证据

- repo_clone_verified: true
- repo_inspection_verified: true
- repo_commit: `5bf4a24104534f241b8245d84ded90c34b852c36`
- inspected_files: `pyproject.toml`, `README.md`, `docs/examples.md`, `docs/contractspec.md`, `docs/api.md`, `docs/benchmarks.md`, `docs/integrations.md`, `docs/getting-started.md`, `docs/index.md`, `docs/metrics.md`, `docs/contracts-catalog.md`, `examples/04_sprt_certification.py`, `examples/03_drift_detection.py`, `examples/05_langgraph_middleware.py`, `examples/07_composition_pipeline.py`, `examples/02_ecommerce_session.py`, `examples/08_mcp_tool_monitoring.py`, `examples/06_crewai_integration.py`, `examples/01_basic_monitoring.py`, `src/agentassert_abc/models.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: 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:1204053098 | https://github.com/qualixar/agentassert-abc | 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:1204053098 | https://github.com/qualixar/agentassert-abc | 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 | github_repo:1204053098 | https://github.com/qualixar/agentassert-abc | 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
- Why it matters: 风险会影响是否适合普通用户安装。
- Evidence: risks.scoring_risks | github_repo:1204053098 | https://github.com/qualixar/agentassert-abc | 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:1204053098 | https://github.com/qualixar/agentassert-abc | 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:1204053098 | https://github.com/qualixar/agentassert-abc | 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。