# agent-powerups - 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 agent-powerups 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_0004` supported 0.86
- **希望把专业流程带进宿主 AI 的用户**：仓库包含 Skill 文档。 证据：`plugins/agent-evaluation-lab/skills/prompt-evaluation-runner/SKILL.md`, `plugins/agent-evaluation-lab/skills/red-team-eval-authoring/SKILL.md`, `plugins/agent-evaluation-lab/skills/skill-evaluation-workbench/SKILL.md`, `plugins/agentic-systems/skills/agent-harness-design/SKILL.md` 等 Claim：`clm_0005` supported 0.86

## 它能做什么

- **AI Skill / Agent 指令资产库**（可做安装前预览）：项目包含可被宿主 AI 读取的 Skill 或 Agent 指令文件，可用于把专业流程带入 Claude、Codex、Cursor 等宿主。 证据：`plugins/agent-evaluation-lab/skills/prompt-evaluation-runner/SKILL.md`, `plugins/agent-evaluation-lab/skills/red-team-eval-authoring/SKILL.md`, `plugins/agent-evaluation-lab/skills/skill-evaluation-workbench/SKILL.md`, `plugins/agentic-systems/skills/agent-harness-design/SKILL.md` 等 Claim：`clm_0001` supported 0.86
- **多宿主安装与分发**（需要安装后验证）：项目包含插件或 marketplace 配置，说明它面向一个或多个 AI 宿主的安装和分发。 证据：`.claude-plugin/marketplace.json`, `.codex-plugin/marketplace.json`, `plugins/agent-evaluation-lab/.claude-plugin/plugin.json`, `plugins/agent-evaluation-lab/.codex-plugin/plugin.json` 等 Claim：`clm_0002` unverified 0.25
- **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_0003` supported 0.86

## 怎么开始

- `git clone https://github.com/yeaight7/agent-powerups.git` 证据：`README.md` Claim：`clm_0006` supported 0.86

## 继续前判断卡

- **当前建议**：先做权限沙盒试用
- **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**：先做权限沙盒试用
- **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**：工具权限边界不能在安装前相信。
- **继续会触碰**：命令执行、宿主 AI 配置、本地环境或项目文件

### 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_0004` supported 0.86
- **适合人群线索：希望把专业流程带进宿主 AI 的用户**（supported）：有 supported claim 或项目证据支撑，但仍不等于真实安装效果。 证据：`plugins/agent-evaluation-lab/skills/prompt-evaluation-runner/SKILL.md`, `plugins/agent-evaluation-lab/skills/red-team-eval-authoring/SKILL.md`, `plugins/agent-evaluation-lab/skills/skill-evaluation-workbench/SKILL.md`, `plugins/agentic-systems/skills/agent-harness-design/SKILL.md` 等 Claim：`clm_0005` supported 0.86
- **能力存在：AI Skill / Agent 指令资产库**（supported）：可以相信项目包含这类能力线索；是否适合你的具体任务仍要试用或安装后验证。 证据：`plugins/agent-evaluation-lab/skills/prompt-evaluation-runner/SKILL.md`, `plugins/agent-evaluation-lab/skills/red-team-eval-authoring/SKILL.md`, `plugins/agent-evaluation-lab/skills/skill-evaluation-workbench/SKILL.md`, `plugins/agentic-systems/skills/agent-harness-design/SKILL.md` 等 Claim：`clm_0001` supported 0.86
- **能力存在：命令行启动或安装流程**（supported）：可以相信项目包含这类能力线索；是否适合你的具体任务仍要试用或安装后验证。 证据：`README.md` Claim：`clm_0003` 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_0006` supported 0.86

### 现在还不能相信

- **工具权限边界不能在安装前相信。**（unverified）：MCP/tool 类项目通常会触碰文件、网络、浏览器或外部 API，必须真实检查权限和日志。
- **真实输出质量不能在安装前相信。**（unverified）：Prompt Preview 只能展示引导方式，不能证明真实项目中的结果质量。
- **宿主 AI 版本兼容性不能在安装前相信。**（unverified）：Claude、Cursor、Codex、Gemini 等宿主加载规则和版本差异必须在真实环境验证。
- **不会污染现有宿主 AI 行为，不能直接相信。**（inferred）：Skill、plugin、AGENTS/CLAUDE/GEMINI 指令可能改变宿主 AI 的默认行为。 证据：`.claude-plugin/marketplace.json`, `.codex-plugin/marketplace.json`, `AGENTS.md`, `CLAUDE.md` 等
- **可安全回滚不能默认相信。**（unverified）：除非项目明确提供卸载和恢复说明，否则必须先在隔离环境验证。
- **真实安装后是否与用户当前宿主 AI 版本兼容？**（unverified）：兼容性只能通过实际宿主环境验证。 证据：`.claude-plugin/marketplace.json`, `.codex-plugin/marketplace.json`, `plugins/agent-evaluation-lab/.claude-plugin/plugin.json`, `plugins/agent-evaluation-lab/.codex-plugin/plugin.json` 等
- **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`
- **宿主 AI 配置**：Claude/Codex/Cursor/Gemini/OpenCode 等宿主的 plugin、Skill 或规则加载配置。 原因：宿主配置会改变 AI 后续工作方式，可能和用户已有规则冲突。 证据：`.claude-plugin/marketplace.json`, `.codex-plugin/marketplace.json`, `AGENTS.md`, `CLAUDE.md` 等
- **本地环境或项目文件**：安装结果、插件缓存、项目配置或本地依赖目录。 原因：安装前无法证明写入范围和回滚方式，需要隔离验证。 证据：`.claude-plugin/marketplace.json`, `.codex-plugin/marketplace.json`, `README.md`, `plugins/agent-evaluation-lab/.claude-plugin/plugin.json` 等
- **宿主 AI 上下文**：AI Context Pack、Prompt Preview、Skill 路由、风险规则和项目事实。 原因：导入上下文会影响宿主 AI 后续判断，必须避免把未验证项包装成事实。

### 最小安全下一步

- **先跑 Prompt Preview**：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。）
- **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。
- **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
- **宿主 AI 插件或 Skill 规则冲突**：新规则可能改变用户现有宿主 AI 的工作方式。 处理方式：安装前先检查插件 manifest 和 Skill 文件，必要时隔离测试。 证据：`.claude-plugin/marketplace.json`, `.codex-plugin/marketplace.json`, `plugins/agent-evaluation-lab/.claude-plugin/plugin.json`, `plugins/agent-evaluation-lab/.codex-plugin/plugin.json` 等 Claim：`clm_0008` 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_0009` supported 0.86
- **待确认**：真实安装后是否与用户当前宿主 AI 版本兼容？。原因：兼容性只能通过实际宿主环境验证。
- **待确认**：项目输出质量是否满足用户具体任务？。原因：安装前预览只能展示流程和边界，不能替代真实评测。
- **待确认**：安装命令是否需要网络、权限或全局写入？。原因：这影响企业环境和个人环境的安装风险。

## 开工前工作上下文

### 加载顺序

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

### 任务路由

- **AI Skill / Agent 指令资产库**：先基于 role_skill_index / evidence_index 帮用户挑选可用角色、Skill 或工作流。 边界：可做安装前 Prompt 体验。 证据：`plugins/agent-evaluation-lab/skills/prompt-evaluation-runner/SKILL.md`, `plugins/agent-evaluation-lab/skills/red-team-eval-authoring/SKILL.md`, `plugins/agent-evaluation-lab/skills/skill-evaluation-workbench/SKILL.md`, `plugins/agentic-systems/skills/agent-harness-design/SKILL.md` 等 Claim：`clm_0001` supported 0.86
- **多宿主安装与分发**：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：`.claude-plugin/marketplace.json`, `.codex-plugin/marketplace.json`, `plugins/agent-evaluation-lab/.claude-plugin/plugin.json`, `plugins/agent-evaluation-lab/.codex-plugin/plugin.json` 等
- **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_0003` supported 0.86

### 上下文规模

- 文件总数：709
- 重要文件覆盖：40/709
- 证据索引条目：180
- 角色 / Skill 条目：203

### 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 agent-powerups 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 agent-powerups 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 agent-powerups 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 203 个角色 / Skill / 项目文档条目。

- **prompt-evaluation-runner**（skill）：Use when evaluating prompts, LLM outputs, red-team suites, or model behavior with local eval configs and safe provider/cost controls. 激活提示：当用户任务与“prompt-evaluation-runner”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/agent-evaluation-lab/skills/prompt-evaluation-runner/SKILL.md`
- **red-team-eval-authoring**（skill）：Use when creating or reviewing red-team eval plugins, attack templates, grader rubrics, safety fixtures, or model-risk test metadata. 激活提示：当用户任务与“red-team-eval-authoring”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/agent-evaluation-lab/skills/red-team-eval-authoring/SKILL.md`
- **skill-evaluation-workbench**（skill）：Use when designing, running, debugging, or hardening deterministic eval suites for agent skills, prompts, tool workflows, or MCP-backed cases. 激活提示：当用户任务与“skill-evaluation-workbench”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/agent-evaluation-lab/skills/skill-evaluation-workbench/SKILL.md`
- **agent-harness-design**（skill）：Design agent tool sets with stable names, narrow schemas, deterministic output shapes, and explicit error paths. No catch-all tools unless unavoidable. 激活提示：当用户任务与“agent-harness-design”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/agentic-systems/skills/agent-harness-design/SKILL.md`
- **canonical-advisor-routing**（skill）：Process-first advisor routing with artifact capture 激活提示：当用户任务与“canonical-advisor-routing”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/agentic-systems/skills/canonical-advisor-routing/SKILL.md`
- **context-retrieval-loop**（skill）：Deterministic 3-cycle loop for gathering codebase context before acting. Broad search → exact source and tests → target-specific docs and setup. Stop after enough context or report what is still missing. 激活提示：当用户任务与“context-retrieval-loop”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/agentic-systems/skills/context-retrieval-loop/SKILL.md`
- **model-routing**（skill）：Vendor-neutral routing guide for choosing the right model tier by task type. Mechanical work uses a smaller/faster model; implementation uses a standard model; architecture, security, and release audit use the most capable model. 激活提示：当用户任务与“model-routing”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/agentic-systems/skills/model-routing/SKILL.md`
- **tri-model-review**（skill）：Multi-model orchestration — route to two external advisors, then synthesize 激活提示：当用户任务与“tri-model-review”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/agentic-systems/skills/tri-model-review/SKILL.md`
- **context-retrieval-loop**（skill）：Deterministic 3-cycle loop for gathering codebase context before acting. Broad search → exact source and tests → target-specific docs and setup. Stop after enough context or report what is still missing. 激活提示：当用户任务与“context-retrieval-loop”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/codebase-intelligence/skills/context-retrieval-loop/SKILL.md`
- **local-rag-mcp**（skill）：Use when querying, ingesting, or maintaining a local RAG MCP corpus for semantic document retrieval with privacy controls. 激活提示：当用户任务与“local-rag-mcp”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/codebase-intelligence/skills/local-rag-mcp/SKILL.md`
- **managed-codebase-context**（skill）：Use when connecting to a managed codebase-context MCP/session service, checking stale maps, or safely using MCP-provided repository context. 激活提示：当用户任务与“managed-codebase-context”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/codebase-intelligence/skills/managed-codebase-context/SKILL.md`
- **search-before-building**（skill）：Check existing repo capability, external libraries, MCP options, and maintenance risk before writing custom code. Decide adopt/wrap/build with explicit criteria. 激活提示：当用户任务与“search-before-building”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/codebase-intelligence/skills/search-before-building/SKILL.md`
- **structured-code-search-mcp**（skill）：Use when designing or using MCP-backed structured code search with search, AST query, symbol inventory, and bounded extraction workflows. 激活提示：当用户任务与“structured-code-search-mcp”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/codebase-intelligence/skills/structured-code-search-mcp/SKILL.md`
- **ai-slop-cleaner**（skill）：Run an anti-slop cleanup workflow on AI-generated code — regression-tests-first, smell-by-smell, behavior preserved. 激活提示：当用户任务与“ai-slop-cleaner”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/codebase-maintenance/skills/ai-slop-cleaner/SKILL.md`
- **architecture-simplification**（skill）：Use to collapse over-engineered abstractions, remove unnecessary layers, or consolidate redundant logic. 激活提示：当用户任务与“architecture-simplification”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/codebase-maintenance/skills/architecture-simplification/SKILL.md`
- **context-retrieval-loop**（skill）：Deterministic 3-cycle loop for gathering codebase context before refactoring or maintenance work. Broad search → exact source and tests → config and build setup. 激活提示：当用户任务与“context-retrieval-loop”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/codebase-maintenance/skills/context-retrieval-loop/SKILL.md`
- **dead-code-removal**（skill）：Use to identify and safely delete unused functions, classes, exports, and files. 激活提示：当用户任务与“dead-code-removal”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/codebase-maintenance/skills/dead-code-removal/SKILL.md`
- **dependency-cleanup**（skill）：Use to audit and remove unused or redundant third-party dependencies from package manifests. 激活提示：当用户任务与“dependency-cleanup”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/codebase-maintenance/skills/dependency-cleanup/SKILL.md`
- **incremental-migration**（skill）：Use when migrating APIs, libraries, or patterns across a large codebase. Ensures safe, step-by-step progress rather than risky mega-commits. 激活提示：当用户任务与“incremental-migration”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/codebase-maintenance/skills/incremental-migration/SKILL.md`
- **naming-and-structure-cleanup**（skill）：Use to enforce consistent naming conventions and file structures across a project without changing business logic. 激活提示：当用户任务与“naming-and-structure-cleanup”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/codebase-maintenance/skills/naming-and-structure-cleanup/SKILL.md`
- **safe-refactor**（skill）：Use when code needs restructuring without changing observable behavior. 激活提示：当用户任务与“safe-refactor”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/codebase-maintenance/skills/safe-refactor/SKILL.md`
- **test-preserving-refactor**（skill）：Use to restructure code while guaranteeing that all existing tests continue to pass. 激活提示：当用户任务与“test-preserving-refactor”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/codebase-maintenance/skills/test-preserving-refactor/SKILL.md`
- **deploy-pipeline-runbook**（skill）：Coordinate multi-system deployment steps as a review-first runbook with explicit approval before any external write or promotion action. 激活提示：当用户任务与“deploy-pipeline-runbook”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/connected-apps/skills/deploy-pipeline-runbook/SKILL.md`
- **bigquery-cost-audit**（skill）：Analyze BigQuery usage, identify cost hotspots, repeated failures, and practical optimization opportunities. 激活提示：当用户任务与“bigquery-cost-audit”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/data-engineering/skills/bigquery-cost-audit/SKILL.md`
- **data-quality**（skill）：Use when adding or reviewing data quality tests for dbt models in warehouse-backed analytics projects. Covers dbt generic tests, singular tests assert .sql , accepted-values macros, dbt utils patterns, cross-system consistency tests, and warehouse-oriented validation. Use when writing data tests, creating assert .sql files, testing business logic, or validating referential integrity. 激活提示：当用户任务与“data-quality”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/data-engineering/skills/data-quality/SKILL.md`
- **dbt-incremental-strategy-audit**（skill）：Audit whether a dbt incremental model uses the right incremental strategy for the repo, the data shape, and the operational constraints. 激活提示：当用户任务与“dbt-incremental-strategy-audit”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/data-engineering/skills/dbt-incremental-strategy-audit/SKILL.md`
- **dbt-preflight**（skill）：Inspect changed dbt assets, estimate blast radius, identify missing tests, and recommend the narrowest safe validation plan. 激活提示：当用户任务与“dbt-preflight”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/data-engineering/skills/dbt-preflight/SKILL.md`
- **dbt-strategy**（skill）：Use when creating or modifying dimensional dbt models in warehouse-backed analytics projects. Covers a four-layer warehouse architecture sources/staging/core/marts , naming conventions, no-alias SQL rule, surrogate-key and missing-record patterns, incremental strategies, deduplication, and common project macros. Use when building fact tables, dimension tables, staging models, writing SQL, or designing tests. 激活提示：当用户任务与“dbt-strategy”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/data-engineering/skills/dbt-strategy/SKILL.md`
- **metric-impact-analyzer**（skill）：Evaluate metric and semantic model changes for BI/reporting breakage and business meaning drift. 激活提示：当用户任务与“metric-impact-analyzer”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/data-engineering/skills/metric-impact-analyzer/SKILL.md`
- **semantic-layer-change-review**（skill）：Use when modifying dbt metrics or semantic models to ensure mathematical correctness and backwards compatibility. 激活提示：当用户任务与“semantic-layer-change-review”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/data-engineering/skills/semantic-layer-change-review/SKILL.md`
- **sql-business-logic-review**（skill）：Review SQL for business logic correctness, semantic drift, aggregation risk, and silent definition changes. 激活提示：当用户任务与“sql-business-logic-review”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/data-engineering/skills/sql-business-logic-review/SKILL.md`
- **bug-hunt**（skill）：Use when reproducing, isolating, and fixing a bug with the smallest safe change. 激活提示：当用户任务与“bug-hunt”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/debugging-diagnostics/skills/bug-hunt/SKILL.md`
- **failure-triage**（skill）：Use when confronted with an unknown failure in CI or production to rapidly categorize the issue before deep debugging. 激活提示：当用户任务与“failure-triage”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/debugging-diagnostics/skills/failure-triage/SKILL.md`
- **flaky-test-investigation**（skill）：Use to diagnose tests that pass and fail intermittently without code changes. 激活提示：当用户任务与“flaky-test-investigation”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/debugging-diagnostics/skills/flaky-test-investigation/SKILL.md`
- **incident-readout**（skill）：Use after fixing a bug to generate a blameless post-mortem summary for human review. 激活提示：当用户任务与“incident-readout”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/debugging-diagnostics/skills/incident-readout/SKILL.md`
- **log-driven-diagnosis**（skill）：Use when debugging complex runtime failures, distributed systems, or issues where a local debugger cannot be attached. 激活提示：当用户任务与“log-driven-diagnosis”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/debugging-diagnostics/skills/log-driven-diagnosis/SKILL.md`
- **minimal-reproduction**（skill）：Use to isolate a bug from a large application into a standalone, runnable script or single test case. 激活提示：当用户任务与“minimal-reproduction”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/debugging-diagnostics/skills/minimal-reproduction/SKILL.md`
- **regression-bisecting**（skill）：Use when a bug was recently introduced but you don't know which commit caused it. 激活提示：当用户任务与“regression-bisecting”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/debugging-diagnostics/skills/regression-bisecting/SKILL.md`
- **systematic-debugging**（skill）：Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes. 激活提示：当用户任务与“systematic-debugging”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/debugging-diagnostics/skills/systematic-debugging/SKILL.md`
- **agent-harness-design**（skill）：Design agent tool sets with stable names, narrow schemas, deterministic output shapes, and explicit error paths. No catch-all tools unless unavoidable. 激活提示：当用户任务与“agent-harness-design”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/dev-vitals/skills/agent-harness-design/SKILL.md`
- **agent-runtime-patterns**（skill）：Use when optimizing agent runtime loops, card packs, MCP session lifecycle, tool-call count, or multi-agent orchestration patterns. 激活提示：当用户任务与“agent-runtime-patterns”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/dev-vitals/skills/agent-runtime-patterns/SKILL.md`
- **agent-session-forensics**（skill）：Use when diagnosing agent session history, interrupted tool loops, missing tool results, timing bottlenecks, or subagent trace correlation. 激活提示：当用户任务与“agent-session-forensics”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/dev-vitals/skills/agent-session-forensics/SKILL.md`
- **brainstorming**（skill）：You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation. 激活提示：当用户任务与“brainstorming”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/dev-vitals/skills/brainstorming/SKILL.md`
- **context-compression**（skill）：Use when conversation context is bloated, constraints are being forgotten, or a compact handoff is needed before continuing work. 激活提示：当用户任务与“context-compression”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/dev-vitals/skills/context-compression/SKILL.md`
- **context-minimization**（skill）：Use continuously during long tasks. Teaches how to read less, output less, and keep the LLM context window lean and fast. 激活提示：当用户任务与“context-minimization”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/dev-vitals/skills/context-minimization/SKILL.md`
- **context-retrieval-loop**（skill）：Deterministic 3-cycle loop for gathering codebase context before acting. Broad search → exact source and tests → target-specific docs and setup. Stop after enough context or report what is still missing. 激活提示：当用户任务与“context-retrieval-loop”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/dev-vitals/skills/context-retrieval-loop/SKILL.md`
- **dispatching-parallel-agents**（skill）：Use when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies 激活提示：当用户任务与“dispatching-parallel-agents”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/dev-vitals/skills/dispatching-parallel-agents/SKILL.md`
- **handoff-discipline**（skill）：Use when completing a task or running out of context limit. Ensures the next session or human engineer has exactly what they need to resume work instantly. 激活提示：当用户任务与“handoff-discipline”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/dev-vitals/skills/handoff-discipline/SKILL.md`
- **no-fluff**（skill）：Ultra-compressed communication mode. Cuts token usage ~75% by dropping filler, articles, and pleasantries while keeping full technical accuracy. Use when user says "no fluff", "be concise", "use less tokens", or similar. 激活提示：当用户任务与“no-fluff”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/dev-vitals/skills/no-fluff/SKILL.md`
- **repo-map**（skill）：Use when the task is to understand an unfamiliar codebase, locate key entry points, or summarize architecture before editing. 激活提示：当用户任务与“repo-map”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/dev-vitals/skills/repo-map/SKILL.md`
- **strategic-context-compaction**（skill）：Compact context at logical phase boundaries — after research, after planning, after debugging — rather than mid-task. Preserves useful state while clearing noise. 激活提示：当用户任务与“strategic-context-compaction”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/dev-vitals/skills/strategic-context-compaction/SKILL.md`
- **task-intake**（skill）：Use at the beginning of a new task. Ensures you fully understand the requirements, boundaries, and acceptance criteria before writing code. 激活提示：当用户任务与“task-intake”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/dev-vitals/skills/task-intake/SKILL.md`
- **using-powerups**（skill）：Use when starting work in a repository with Agent Powerups installed, when a task may match a reusable local skill, command, workflow, hook recipe, AGENTS.md template, or MCP feature. 激活提示：当用户任务与“using-powerups”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/dev-vitals/skills/using-powerups/SKILL.md`
- **verification-before-completion**（skill）：Use when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always 激活提示：当用户任务与“verification-before-completion”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/dev-vitals/skills/verification-before-completion/SKILL.md`
- **writing-plans**（skill）：Use when you have a spec or requirements for a multi-step task, before touching code. 激活提示：当用户任务与“writing-plans”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/dev-vitals/skills/writing-plans/SKILL.md`
- **agent-readable-docs**（skill）：Use when writing technical documentation that needs to be readable by both humans and AI models, converting existing docs to HADS format, validating a HADS document, or optimizing documentation for token-efficient AI consumption. 激活提示：当用户任务与“agent-readable-docs”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/documentation-systems/skills/agent-readable-docs/SKILL.md`
- **api-doc-review**（skill）：Verify that API endpoints match their OpenAPI/Swagger specifications. 激活提示：当用户任务与“api-doc-review”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/documentation-systems/skills/api-doc-review/SKILL.md`
- **architecture-decision-records**（skill）：Record why an architectural choice was made to prevent agents or humans from unintentionally reverting it. 激活提示：当用户任务与“architecture-decision-records”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/documentation-systems/skills/architecture-decision-records/SKILL.md`
- **context-docs**（skill）：Maintain short, focused Markdown files per subsystem to provide agents with isolated context. 激活提示：当用户任务与“context-docs”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/documentation-systems/skills/context-docs/SKILL.md`
- **context-retrieval-loop**（skill）：Deterministic 3-cycle loop for gathering documentation context before writing or updating docs. Broad search → exact source and existing docs → project conventions and setup. 激活提示：当用户任务与“context-retrieval-loop”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/documentation-systems/skills/context-retrieval-loop/SKILL.md`
- **doc-consistency-check**（skill）：Audit documentation for broken file paths, outdated commands, and renamed variables. 激活提示：当用户任务与“doc-consistency-check”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/documentation-systems/skills/doc-consistency-check/SKILL.md`
- **handoff-documentation**（skill）：Write state-restoration documents for passing tasks between agents or engineers. 激活提示：当用户任务与“handoff-documentation”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/documentation-systems/skills/handoff-documentation/SKILL.md`
- **readme-hardening**（skill）：Ensure the project README provides immediate, exact commands for setup, testing, and deployment to help agents and humans bootstrap quickly. 激活提示：当用户任务与“readme-hardening”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/documentation-systems/skills/readme-hardening/SKILL.md`
- **review-comment-style-mining**（skill）：Use when mining PR review comments and text diffs for reusable writing, documentation, tone, and editorial improvement patterns. 激活提示：当用户任务与“review-comment-style-mining”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/documentation-systems/skills/review-comment-style-mining/SKILL.md`
- **gh-address-comments**（skill）：Address actionable GitHub pull request review feedback. Use when the user wants to inspect unresolved review threads, requested changes, or inline review comments on a PR, then implement selected fixes. 激活提示：当用户任务与“gh-address-comments”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/github-ops/skills/gh-address-comments/SKILL.md`
- **github-ci-failure-triage**（skill）：Inspect GitHub PR checks, fetch actionable failure logs, summarize the breakage, and propose a local fix plan before changing code. 激活提示：当用户任务与“github-ci-failure-triage”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/github-ops/skills/github-ci-failure-triage/SKILL.md`
- **pr-review-ci-loop**（skill）：Run a review and CI loop around a pull request with explicit approval gates for code changes, remote writes, and follow-up actions. 激活提示：当用户任务与“pr-review-ci-loop”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/github-ops/skills/pr-review-ci-loop/SKILL.md`
- **baseline-comparison-review**（skill）：Ensure that new complex models actually outperform simple, naive baselines. 激活提示：当用户任务与“baseline-comparison-review”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/machine-learning-ops/skills/baseline-comparison-review/SKILL.md`
- **dataset-split-review**（skill）：Audit the methodology used to split data into train, validation, and test sets. 激活提示：当用户任务与“dataset-split-review”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/machine-learning-ops/skills/dataset-split-review/SKILL.md`
- **experiment-tracking-review**（skill）：Verify that all hyperparameters, metrics, and data references are properly logged. 激活提示：当用户任务与“experiment-tracking-review”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/machine-learning-ops/skills/experiment-tracking-review/SKILL.md`
- **ml-leakage-check**（skill）：Identify and prevent target leakage in ML preprocessing pipelines. 激活提示：当用户任务与“ml-leakage-check”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/machine-learning-ops/skills/ml-leakage-check/SKILL.md`
- **model-evaluation-reporting**（skill）：Standardize the reporting of model metrics to ensure statistical rigor and business relevance. 激活提示：当用户任务与“model-evaluation-reporting”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/machine-learning-ops/skills/model-evaluation-reporting/SKILL.md`
- **reproducible-training-runs**（skill）：Analyzes ML training scripts to enforce seed setting, deterministic operations, and environment tracking for exact reproducibility. 激活提示：当用户任务与“reproducible-training-runs”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/machine-learning-ops/skills/reproducible-training-runs/SKILL.md`
- **training-pipeline-debugging**（skill）：Diagnose NaN losses, out-of-memory errors, and shape mismatches in deep learning or ML pipelines. 激活提示：当用户任务与“training-pipeline-debugging”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/machine-learning-ops/skills/training-pipeline-debugging/SKILL.md`
- **mcp-server-builder**（skill）：Design high-quality MCP servers around workflows, narrow schemas, context-aware outputs, and actionable errors. Use when building or reviewing MCP tools for real agent tasks. 激活提示：当用户任务与“mcp-server-builder”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/mcp-development/skills/mcp-server-builder/SKILL.md`
- **defuddle**（skill）：Use when the user provides a URL to a standard web page and clean Markdown extraction with Defuddle would reduce clutter and token cost. 激活提示：当用户任务与“defuddle”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/memory-optimization/skills/defuddle/SKILL.md`
- **graphify**（skill）：any input code, docs, papers, images → knowledge graph → clustered communities → HTML + JSON + audit report. Use when the user wants to build, inspect, or query a cross-file knowledge graph from code, docs, papers, images, or videos with upstream graphify. 激活提示：当用户任务与“graphify”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/memory-optimization/skills/graphify/SKILL.md`
- **markitdown-file-intake**（skill）：Use when the user provides a MarkItDown-supported file or URL and converting it to Markdown first will make inspection easier, cheaper, or more reliable. 激活提示：当用户任务与“markitdown-file-intake”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/memory-optimization/skills/markitdown-file-intake/SKILL.md`
- **memory-build-workflow**（skill）：Use when a user needs to build or refresh persistent graph memory from a mixed corpus and the right path may include graphify, incremental update, or helper conversion before ingestion. 激活提示：当用户任务与“memory-build-workflow”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/memory-optimization/skills/memory-build-workflow/SKILL.md`
- **memory-optimization-workflow**（skill）：Use when deciding the lowest-cost context path for a mixed corpus, especially when choosing among direct reading, helper conversion, graph build, graph update, or graph query. 激活提示：当用户任务与“memory-optimization-workflow”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`plugins/memory-optimization/skills/memory-optimization-workflow/SKILL.md`
- 其余 123 个条目见 `AI_CONTEXT_PACK.json`。

## 证据索引

- 共索引 180 条证据。

- **Gemini Setup**（documentation）：Status: compatibility and agent-curated setup path. For manual install, prefer apx install gemini or apx install gemini --full . This document describes what apx setup gemini still does; it is not an official Gemini CLI integration claim. 证据：`docs/setup/gemini.md`
- **Repository Guidelines**（documentation）：Project Structure & Module Organization Agent Powerups is a Node.js/TypeScript CLI plus a catalog of agent assets. Core CLI code lives in src/cli/ , with command handlers in src/cli/commands/ and shared helpers in src/cli/utils/ . Tests live in test/ as .test.ts . 证据：`AGENTS.md`
- **CLAUDE.md**（documentation）：This file provides guidance to Claude Code claude.ai/code when working with code in this repository. 证据：`CLAUDE.md`
- **Project Overview**（documentation）：Agent Powerups is an "Oh My Zsh-style" collection of reusable skills, slash commands, MCP configs, hooks, AGENTS.md templates, and workflows for coding agents. It provides a local-first CLI tool apx for browsing, validating, running, and explicitly installing these agent powerups. 证据：`GEMINI.md`
- **What Is Here**（documentation）：Quickstart · Plugin Bundles · Installation · Security Model · Contributing 证据：`README.md`
- **Plugins**（documentation）：Domain-specific plugin bundles for Agent Powerups. Each plugin extends the base powerups with deeper, specialized skills, agents, and commands for a particular engineering domain. 证据：`plugins/README.md`
- **AGENTS.md**（documentation）：- Inspect model lineage before changing SQL. - Keep metric and semantic changes explicit. - Do not change warehouse credentials or profiles. 证据：`agents-md/dbt-project/AGENTS.md`
- **AGENTS.md**（documentation）：- Preserve seeds, splits, metrics, and experiment tracking unless requested. - Treat data changes as high-risk. - Keep model comparisons reproducible. 证据：`agents-md/ml-project/AGENTS.md`
- **AGENTS.md**（documentation）：- Preserve public APIs and contributor workflows. - Read issue or PR context before editing. - Keep changelog, docs, and tests aligned with user-visible changes. 证据：`agents-md/open-source-maintainer/AGENTS.md`
- **AGENTS.md**（documentation）：- Read package metadata and tests before editing. - Prefer small API-compatible changes. - Do not add dependencies or publishing config unless requested. 证据：`agents-md/python-library/AGENTS.md`
- **AGENTS.md**（documentation）：- Read the smallest relevant set of files before editing. - Prefer minimal diffs. - Preserve the existing package manager, lint, test, and build flows. - Run the narrowest meaningful validation before claiming completion. - Do not add dependencies, secrets, or global config changes unless explicitly requested. 证据：`agents-md/typescript-app/AGENTS.md`
- **Claude Code Setup Example**（documentation）：Goal: stage Agent Powerups for Claude Code review. 证据：`examples/claude-code/README.md`
- **Codex Setup Example**（documentation）：Goal: stage Agent Powerups for Codex review. 证据：`examples/codex/README.md`
- **Minimal Setup Example**（documentation）：Goal: inspect Agent Powerups without mutating an agent's global config. 证据：`examples/minimal/README.md`
- **agent-evaluation-lab**（documentation）：Prompt, skill, red-team, and agent behavior evaluation workflows. 证据：`plugins/agent-evaluation-lab/GEMINI.md`
- **agentic-systems**（documentation）：Multi-model orchestration and advisor routing: tri-model review and canonical advisor routing. 证据：`plugins/agentic-systems/GEMINI.md`
- **Agents**（documentation）：Agents Experimental agentic-systems agents. 证据：`plugins/agentic-systems/agents/README.md`
- **Commands**（documentation）：Commands Experimental agentic-systems commands. 证据：`plugins/agentic-systems/commands/README.md`
- **codebase-intelligence**（documentation）：Search-first codebase understanding, context retrieval, mapping, pattern detection, and living project intelligence. 证据：`plugins/codebase-intelligence/GEMINI.md`
- **codebase-maintenance**（documentation）：Safe refactoring, dead-code removal, legacy modernization, and dependency hygiene. 证据：`plugins/codebase-maintenance/GEMINI.md`
- **connected-apps**（documentation）：Review-first runbooks for connected-app workflows and multi-system deployment sequencing. 证据：`plugins/connected-apps/GEMINI.md`
- **context-efficiency**（documentation）：Context-efficient routing: dispatch tables for workflow, review, and codebase-intelligence commands. 证据：`plugins/context-efficiency/GEMINI.md`
- **data-engineering**（documentation）：Analytics engineering, dbt patterns, BigQuery optimization, data quality, and pipeline workflows. 证据：`plugins/data-engineering/GEMINI.md`
- **debugging-diagnostics**（documentation）：Error diagnosis, fault isolation, log forensics, and incident investigation. 证据：`plugins/debugging-diagnostics/GEMINI.md`
- **dev-vitals**（documentation）：Core developer workflow skills: planning, execution, communication, and powerup discovery. 证据：`plugins/dev-vitals/GEMINI.md`
- **documentation-systems**（documentation）：Documentation architecture, API reference authoring, ADR writing, readme hardening, and doc quality. 证据：`plugins/documentation-systems/GEMINI.md`
- **github-ops**（documentation）：GitHub review, comment handling, and CI triage workflows with approval-gated write actions. 证据：`plugins/github-ops/GEMINI.md`
- **machine-learning-ops**（documentation）：ML pipeline design, experiment tracking, model evaluation, training debugging, and reproducibility. 证据：`plugins/machine-learning-ops/GEMINI.md`
- **mcp-development**（documentation）：Workflow-first MCP server design with schema-first tools, context-aware outputs, and evaluation discipline. 证据：`plugins/mcp-development/GEMINI.md`
- **memory-optimization**（documentation）：Graph-backed memory and context optimization bundle with graphify as the primary engine and document-conversion helpers used only when they reduce noise or reread cost. 证据：`plugins/memory-optimization/GEMINI.md`
- **CLAUDE.md stub**（documentation）：This is a stub reference file. The graphify skill references CLAUDE.md as a documentation source in the target repository. 证据：`plugins/memory-optimization/skills/graphify/references/CLAUDE.md`
- **quality-gates**（documentation）：Quality enforcement: TDD discipline, code review workflows, release verification, and CI failure analysis. 证据：`plugins/quality-gates/GEMINI.md`
- **security-guardrails**（documentation）：Security vulnerability detection: OWASP Top 10 analysis, secrets scanning, dependency audits. 证据：`plugins/security-guardrails/GEMINI.md`
- **Commands**（documentation）：Commands Experimental security-guardrails commands. 证据：`plugins/security-guardrails/commands/README.md`
- **skill-authoring**（documentation）：Skill creation and extraction workflows for reusable agent guidance. 证据：`plugins/skill-authoring/GEMINI.md`
- **software-engineering**（documentation）：Subagent orchestration, wave execution, worktree isolation, persistent completion, autonomous delivery, and migration batching. 证据：`plugins/software-engineering/GEMINI.md`
- **Commands**（documentation）：Commands Experimental software-engineering commands. 证据：`plugins/software-engineering/commands/README.md`
- **spec-driven-development**（documentation）：Requirements clarification, phase planning, workstreams, and wave-based execution for reviewable delivery. 证据：`plugins/spec-driven-development/GEMINI.md`
- **spec-quality-gates**（documentation）：Quality gates for spec-driven development: adversarial plan verification and structured code review. 证据：`plugins/spec-quality-gates/GEMINI.md`
- **tool-integrations**（documentation）：Browser, MCP, retrieval, and agent runtime integration skills with conservative setup and safety boundaries. 证据：`plugins/tool-integrations/GEMINI.md`
- **CLAUDE.md stub**（documentation）：This is a stub reference file. The graphify skill references CLAUDE.md as a documentation source in the target repository. 证据：`skills/graphify/references/CLAUDE.md`
- **Core Agent Tools - Refactored Assets**（documentation）：Core Agent Tools - Refactored Assets 证据：`staging/refactor/core-agent-tools/README.md`
- **Package**（package_manifest）：{ "name": "agent-powerups", "version": "0.3.0", "description": "Local-first CLI for browsing, validating, running, and explicitly writing agent powerups.", "license": "Apache-2.0", "type": "module", "repository": { "type": "git", "url": "git+https://github.com/yeaight7/agent-powerups.git" }, "bugs": { "url": "https://github.com/yeaight7/agent-powerups/issues" }, "homepage": "https://github.com/yeaight7/agent-powerups", "keywords": "claude-code", "codex", "ai-agents", "skills", "mcp", "developer-tools", "cli", "gemini-cli", "gemini", "skill-pack", "plugin" , "bin": { "apx": "dist/cli/apx.js" }, "engines": { "node": " =20" }, "files": "dist/", "!dist/test/", "skills/", "!skills/graphify/refer… 证据：`package.json`
- **Contributing to Agent Powerups**（documentation）：Keep contributions small, explicit, and portable. 证据：`CONTRIBUTING.md`
- **Prompt Evaluation Runner**（skill_instruction）：When to use Use this skill when you need to evaluate an LLM app, test a prompt, or run red-teaming/vulnerability scans against a target model or application. 证据：`plugins/agent-evaluation-lab/skills/prompt-evaluation-runner/SKILL.md`
- **Red-Team Eval Authoring**（skill_instruction）：When To Use - Adding a new red-team plugin or grader. - Editing attack templates, rubric tags, or plugin metadata. - Reviewing multimodal or tool-use safety evals for false positives/negatives. 证据：`plugins/agent-evaluation-lab/skills/red-team-eval-authoring/SKILL.md`
- **Skill Evaluation Workbench**（skill_instruction）：When To Use - A skill or prompt needs repeatable quality checks across models. - A workflow needs file-based graders, command traces, or local artifact checks. - A tool/MCP skill needs a hidden service fixture or sandboxed test workspace. - A previous agent attempt failed and you need trace-driven diagnosis before editing instructions. 证据：`plugins/agent-evaluation-lab/skills/skill-evaluation-workbench/SKILL.md`
- **Agent Harness Design**（skill_instruction）：Use when designing or improving how an agent invokes tools, handles errors, and decides when to stop. 证据：`plugins/agentic-systems/skills/agent-harness-design/SKILL.md`
- **Canonical Advisor Routing**（skill_instruction）：Route a prompt through a local provider CLI and persist the result as an artifact. 证据：`plugins/agentic-systems/skills/canonical-advisor-routing/SKILL.md`
- **Context Retrieval Loop**（skill_instruction）：Gather the right codebase context before making changes or spawning subagents. Prevents acting on incomplete information. 证据：`plugins/agentic-systems/skills/context-retrieval-loop/SKILL.md`
- **Model Routing**（skill_instruction）：Choose the right model tier before starting a task. Overusing a capable model wastes cost and context. Underusing it produces lower quality on complex work. 证据：`plugins/agentic-systems/skills/model-routing/SKILL.md`
- **Tri-Model Review**（skill_instruction）：Tri-model review routes through two external advisor CLIs, then synthesizes both outputs into one answer. 证据：`plugins/agentic-systems/skills/tri-model-review/SKILL.md`
- **Context Retrieval Loop**（skill_instruction）：Gather the right codebase context before making changes or spawning subagents. Prevents acting on incomplete information. 证据：`plugins/codebase-intelligence/skills/context-retrieval-loop/SKILL.md`
- **Local RAG MCP**（skill_instruction）：When to use Use when the task requires semantic search, document ingestion, or querying a local vector database for context retrieval, and an appropriate MCP server is available. 证据：`plugins/codebase-intelligence/skills/local-rag-mcp/SKILL.md`
- **Managed Codebase Context**（skill_instruction）：When to use Use when managing persistent sessions, dynamic codebase mappings, or managed context states using a dedicated codebase-context MCP server. 证据：`plugins/codebase-intelligence/skills/managed-codebase-context/SKILL.md`
- **Search Before Building**（skill_instruction）：Before implementing new functionality, verify it does not already exist and that the best option has been considered. 证据：`plugins/codebase-intelligence/skills/search-before-building/SKILL.md`
- **Structured Code Search MCP**（skill_instruction）：When to use Use when developing a new MCP server, debugging an MCP connection, or integrating an agent with standard MCP protocols. 证据：`plugins/codebase-intelligence/skills/structured-code-search-mcp/SKILL.md`
- **Purpose**（skill_instruction）：Reduce AI-generated code bloat through systematic, smell-by-smell cleanup that preserves existing behavior and raises signal quality. 证据：`plugins/codebase-maintenance/skills/ai-slop-cleaner/SKILL.md`
- **Architecture Simplification**（skill_instruction）：Over time, codebases accumulate "just in case" abstractions. This skill guides the safe removal of unnecessary complexity. 证据：`plugins/codebase-maintenance/skills/architecture-simplification/SKILL.md`
- **Context Retrieval Loop Codebase Maintenance**（skill_instruction）：Context Retrieval Loop Codebase Maintenance 证据：`plugins/codebase-maintenance/skills/context-retrieval-loop/SKILL.md`
- 其余 120 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/setup/gemini.md`, `AGENTS.md`, `CLAUDE.md`
- **回答用户时区分可预览内容与必须安装后才能验证的内容。**：安装前体验的消费者价值来自降低误装和误判，而不是伪装成真实运行。 证据：`docs/setup/gemini.md`, `AGENTS.md`, `CLAUDE.md`

## 用户开工前应该回答的问题

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

## 验收标准

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

---

## Doramagic Context Augmentation

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

## Human Manual 骨架

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

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

- **项目概览**：importance `high`
  - source_paths: README.md, SECURITY.md, catalog.json, package.json
- **安装指南**：importance `high`
  - source_paths: docs/installation.md, docs/compatibility.md, src/index.ts, tsconfig.json
- **apx CLI 参考**：importance `high`
  - source_paths: src/cli/apx.ts, src/cli/commands, src/cli/utils
- **资产目录系统**：importance `high`
  - source_paths: catalog.json, docs/catalog-schema.md, scripts/validate-skills.py, scripts/validate-catalog.py
- **插件包系统**：importance `high`
  - source_paths: plugins/README.md, plugin-bundles.json, .claude-plugin/marketplace.json, .codex-plugin/marketplace.json, docs/plugin-bundles-schema.json
- **技能系统**：importance `high`
  - source_paths: skills/systematic-debugging/SKILL.md, skills/writing-plans/SKILL.md, skills/using-powerups/SKILL.md, skills/skill-authoring-guide/SKILL.md
- **命令系统**：importance `medium`
  - source_paths: commands/generic, commands/claude-code, commands/codex, commands/bug-check.md, commands/security-audit.md
- **钩子系统**：importance `medium`
  - source_paths: hooks/productivity, hooks/quality, hooks/safety, docs/security-model.md

## Repo Inspection Evidence / 源码检查证据

- repo_clone_verified: true
- repo_inspection_verified: true
- repo_commit: `a73f89c13e78d5fe30f46c7c1a8c03bf350e9429`
- inspected_files: `package.json`, `README.md`, `docs/philosophy.md`, `docs/release-checklist-v0.1.0.md`, `docs/security-model.md`, `docs/roadmap.md`, `docs/release-checklist-v0.1.1.md`, `docs/profiles-schema.md`, `docs/plugin-bundles-schema.json`, `docs/installation.md`, `docs/catalog-schema.md`, `docs/compatibility.md`, `docs/authoring-guide.md`, `docs/tool-requirements.md`, `docs/mcp-configs.md`, `docs/setup/gemini.md`, `docs/setup/codex.md`, `docs/setup/claude-code.md`, `examples/minimal/README.md`, `examples/codex/README.md`

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

## Doramagic Pitfall Constraints / 踩坑约束

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

### Constraint 1: 仓库名和安装名不一致

- Trigger: 仓库名 `agent-powerups` 与安装入口 `markitdown` 不完全一致。
- Host AI rule: 在 npm/PyPI/GitHub 上确认包名映射和官方 README 说明。
- Why it matters: 用户照着仓库名搜索包或照着包名找仓库时容易走错入口。
- Evidence: identity.distribution | github_repo:1222971895 | https://github.com/yeaight7/agent-powerups | repo=agent-powerups; install=markitdown
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 2: 可能修改宿主 AI 配置

- Trigger: 项目面向 Claude/Cursor/Codex/Gemini/OpenCode 等宿主，或安装命令涉及用户配置目录。
- Host AI rule: 列出会写入的配置文件、目录和卸载/回滚步骤。
- Why it matters: 安装可能改变本机 AI 工具行为，用户需要知道写入位置和回滚方法。
- Evidence: capability.host_targets | github_repo:1222971895 | https://github.com/yeaight7/agent-powerups | host_targets=mcp_host, claude, claude_code
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 3: 能力判断依赖假设

- Trigger: README/documentation is current enough for a first validation pass.
- Host AI rule: 将假设转成下游验证清单。
- Why it matters: 假设不成立时，用户拿不到承诺的能力。
- Evidence: capability.assumptions | github_repo:1222971895 | https://github.com/yeaight7/agent-powerups | README/documentation is current enough for a first validation pass.
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 4: 维护活跃度未知

- Trigger: 未记录 last_activity_observed。
- Host AI rule: 补 GitHub 最近 commit、release、issue/PR 响应信号。
- Why it matters: 新项目、停更项目和活跃项目会被混在一起，推荐信任度下降。
- Evidence: evidence.maintainer_signals | github_repo:1222971895 | https://github.com/yeaight7/agent-powerups | last_activity_observed missing
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 5: 下游验证发现风险项

- Trigger: no_demo
- Host AI rule: 进入安全/权限治理复核队列。
- Why it matters: 下游已经要求复核，不能在页面中弱化。
- Evidence: downstream_validation.risk_items | github_repo:1222971895 | https://github.com/yeaight7/agent-powerups | 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
- Host AI rule: 把风险写入边界卡，并确认是否需要人工复核。
- Why it matters: 风险会影响是否适合普通用户安装。
- Evidence: risks.scoring_risks | github_repo:1222971895 | https://github.com/yeaight7/agent-powerups | no_demo; severity=medium
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 7: 来源证据：v0.1.4

- Trigger: GitHub 社区证据显示该项目存在一个安全/权限相关的待验证问题：v0.1.4
- Host AI rule: 来源显示可能已有修复、规避或版本变化，说明书中必须标注适用版本。
- Why it matters: 可能增加新用户试用和生产接入成本。
- Evidence: community_evidence:github | cevd_ee1d355f496c46158442305fd9ed9206 | https://github.com/yeaight7/agent-powerups/releases/tag/v0.1.4 | 来源讨论提到 npm 相关条件，需在安装/试用前复核。
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 8: 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:1222971895 | https://github.com/yeaight7/agent-powerups | 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 9: 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:1222971895 | https://github.com/yeaight7/agent-powerups | 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。