# tandem - 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 tandem 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_0019` 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：`agent-templates/packs/bio-informatics-pack/skills/bio-informatics-analyst/SKILL.md`, `agent-templates/packs/data-visualization-pack/skills/data-visualization-analyst/SKILL.md`, `agent-templates/packs/finance-analysis-pack/skills/finance-analysis-analyst/SKILL.md`, `agent-templates/packs/legal-research-pack/skills/legal-research-analyst/SKILL.md` 等 Claim：`clm_0020` supported 0.86 ## 它能做什么 - **Tandem Engine Runtime**（需要安装后验证）：The headless HTTP+SSE runtime binary that powers all Tandem clients. It owns sessions, runs, memory, tools, approvals, artifacts, and audit records below the model layer. 证据：`engine/README.md`, `packages/tandem-engine/README.md`, `README.md`, `README.md` Claim：`clm_0001` supported 0.86, `clm_0003` supported 0.86, `clm_0006` supported 0.86, `clm_0009` supported 0.86 等 - **Session and Run Management**（需要安装后验证）：Create, list, and manage agent sessions with async run lifecycles. Sessions carry tenant/principal context through runs, context, and provider credentials. 证据：`packages/tandem-client-ts/README.md`, `packages/tandem-client-py/README.md`, `packages/tandem-control-panel/README.md`, `docs/ENGINE_COMMUNICATION.md` Claim：`clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0005` supported 0.86, `clm_0006` supported 0.86 等 - **Tool Execution with Approval Gates**（需要安装后验证）：Built-in and MCP connector tools execute under scoped authority with optional human approval gates for consequential actions before execution completes. 证据：`README.md`, `packages/tandem-tui/README.md`, `packages/tandem-control-panel/README.md`, `docs/ENGINE_CLI.md` 等 Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0006` supported 0.86 等 - **MCP (Model Context Protocol) Integration**（需要安装后验证）：Tandem auto-registers MCP servers including the official GitHub MCP server when a PAT is available. Supports 162+ cataloged MCP servers with security override configuration. 证据：`docs/MCP_IMPROVEMENTS.md`, `crates/tandem-server/resources/mcp-catalog/README.md`, `crates/tandem-server/resources/mcp-catalog/README.md`, `.agents/README.md` Claim：`clm_0004` supported 0.86 - **Browser Automation Tools**（需要安装后验证）：Built-in browser automation for QA workflows including status checks, navigation, clicks, typing, screenshots, and DOM extraction with allowlist-based tool scoping. 证据：`packages/tandem-client-py/README.md`, `packages/tandem-client-py/README.md`, `packages/tandem-enterprise/README.md` Claim：`clm_0002` supported 0.86, `clm_0005` supported 0.86, `clm_0010` supported 0.86, `clm_0011` supported 0.86 等 - **Permissioned Memory Management**（需要安装后验证）：Tenant-aware memory with embeddings for semantic search. Memory is scoped per tenant/principal and supports global cross-project memory as an opt-in feature. 证据：`engine/README.md`, `engine/README.md`, `README.md`, `packages/tandem-control-panel/README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0006` supported 0.86 等 - **Scheduled and On-Demand Routines**（需要安装后验证）：Create cron-scheduled automation routines that execute prompts with scoped tool allowlists, plus immediate run-now triggers and artifact retrieval. 证据：`packages/tandem-client-ts/README.md`, `packages/tandem-client-ts/README.md`, `packages/tandem-control-panel/README.md` Claim：`clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0006` supported 0.86, `clm_0007` supported 0.86 等 - **Workflow Plan Compiler**（需要安装后验证）：High-level goal-to-plan compiler producing governed plan packages, runtime projections, and draft lifecycle behavior. Supports preview, interactive chat refinement, apply, and import flows. 证据：`crates/tandem-plan-compiler/README.md`, `crates/tandem-plan-compiler/README.md`, `packages/tandem-client-ts/README.md` Claim：`clm_0002` supported 0.86, `clm_0007` supported 0.86, `clm_0008` supported 0.86, `clm_0010` supported 0.86 - **Multi-Client Entry Points**（可做安装前预览）：Tandem ships four client surfaces: Tauri desktop app, terminal TUI, web control panel, and editable scaffold. All share the same engine runtime. 证据：`README.md`, `apps/tandem-desktop/package.json`, `crates/tandem-tui/README.md`, `packages/tandem-control-panel/README.md` 等 Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0006` supported 0.86 等 - **Python and TypeScript SDKs**（需要安装后验证）：Language-native SDKs exposing the engine HTTP+SSE API with async/sync support, session management, streaming, tool execution, routines, and workflow plans. 证据：`packages/tandem-client-py/README.md`, `packages/tandem-client-py/README.md`, `packages/tandem-client-ts/README.md`, `packages/tandem-client-ts/package.json` Claim：`clm_0002` supported 0.86, `clm_0005` supported 0.86, `clm_0007` supported 0.86, `clm_0008` supported 0.86 等 - **Multi-Channel Integration**（需要安装后验证）：Built-in channel support for Telegram, Discord, and Slack with registry-based lifecycle, endpoint behavior diagnostics, and channel_uploads for file handling. 证据：`packages/tandem-control-panel/README.md`, `docs/CHANNELS_LIFECYCLE_AND_DIAGNOSTICS.md`, `packages/tandem-client-py/README.md` Claim：`clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0005` supported 0.86, `clm_0006` supported 0.86 等 - **Provider-Agnostic LLM Routing**（可做安装前预览）：Engine supports OpenRouter, Anthropic, OpenAI, OpenCode Zen, Ollama, Groq, Mistral, Together, Cohere, Azure, Vertex, Bedrock, and GitHub Copilot providers via unified API surface. 证据：`README.md`, `packages/tandem-engine/README.md`, `engine/README.md` Claim：`clm_0001` supported 0.86, `clm_0003` supported 0.86, `clm_0006` supported 0.86, `clm_0009` supported 0.86 等 - **Enterprise Linux Binary Distribution**（需要安装后验证）：v0.5.11 ships a dedicated Linux x64 enterprise binary with browser automation and enterprise-full routes compiled in, plus npm wrapper for hosted sidecar deployment. 证据：`packages/tandem-enterprise/README.md`, `packages/tandem-enterprise/package.json` Claim：`clm_0005` supported 0.86, `clm_0013` supported 0.86 - **EU AI Act Compliance Documentation**（可做安装前预览）：Public compliance starter pack with control mapping, Article 50 transparency guidance, Annex IV technical documentation template, deployer instructions, and limitations matrix. 证据：`docs/compliance/README.md`, `docs/EU_AI_ACT_COMPLIANCE.md` Claim：`clm_0014` supported 0.86 - **Concurrent Batch Processing**（需要安装后验证）：Run multiple prompts concurrently with configurable concurrency limits and JSON task specification file input. 证据：`packages/tandem-engine/README.md`, `packages/tandem-engine/README.md` Claim：`clm_0001` supported 0.86, `clm_0012` supported 0.86, `clm_0015` supported 0.86 - **Web Research Tools**（需要安装后验证）：Built-in webfetch and websearch tools for URL content extraction and search queries, with markdown conversion and noise stripping. 证据：`docs/ENGINE_CLI.md`, `docs/ENGINE_CLI.md`, `scripts/bench-js/README.md` Claim：`clm_0003` supported 0.86, `clm_0016` supported 0.86, `clm_0017` supported 0.86 - **Token Generation and API Authentication**（可做安装前预览）：Engine generates and manages API tokens for client authentication. Supports explicit token via env/flag, token file, and unsafe local-only opt-out. 证据：`packages/tandem-client-py/README.md`, `docs/ENGINE_CLI.md`, `docs/ENGINE_COMMUNICATION.md` Claim：`clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0005` supported 0.86, `clm_0010` supported 0.86 等 - **Swarm Orchestration with Live Flow Visualization**（需要安装后验证）：Node-based swarm orchestration for multi-agent coordination with live flow visualization in the control panel. 证据：`packages/tandem-control-panel/README.md`, `docs/ENGINE_RUNTIME.md` Claim：`clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0006` supported 0.86, `clm_0007` supported 0.86 等 ## 怎么开始 - `npm i -g @frumu/tandem` 证据：`README.md` Claim：`clm_0021` supported 0.86, `clm_0027` supported 0.86 - `npm install -g @frumu/tandem` 证据：`README.md` Claim：`clm_0022` supported 0.86, `clm_0028` supported 0.86 - `git clone https://github.com/frumu-ai/tandem.git` 证据：`README.md` Claim：`clm_0023` supported 0.86 - `npx create-tandem-panel my-panel` 证据：`packages/create-tandem-panel/README.md` Claim：`clm_0024` supported 0.86 - `pip install tandem-client` 证据：`packages/tandem-client-py/README.md` Claim：`clm_0025` supported 0.86 - `npm install @frumu/tandem-client` 证据：`packages/tandem-client-ts/README.md` Claim：`clm_0026` supported 0.86 - `npm i -g @frumu/tandem-panel` 证据：`packages/tandem-control-panel/README.md` Claim：`clm_0027` supported 0.86 - `npm install -g @frumu/tandem-tui` 证据：`packages/tandem-tui/README.md` Claim：`clm_0028` supported 0.86 - `curl -s http://127.0.0.1:39731/global/health \` 证据：`guide/src/content/docs/agent-workflow-mission-quickstart.md` Claim：`clm_0029` unverified 0.25 ## 继续前判断卡 - **当前建议**：需要管理员/安全审批 - **为什么**：继续前可能涉及密钥、账号、外部服务或敏感上下文，建议先经过管理员或安全审批。 ### 30 秒判断 - **现在怎么做**：需要管理员/安全审批 - **最小安全下一步**：先跑 Prompt Preview；若涉及凭证或企业环境，先审批再试装 - **先别相信**：研究结论、引用和实验结果不能在安装前相信。 - **继续会触碰**：研究判断、命令执行、宿主 AI 配置 ### 现在可以相信 - **适合人群线索：正在使用 Claude/Codex/Cursor/Gemini 等宿主 AI 的开发者**（supported）：有 supported claim 或项目证据支撑，但仍不等于真实安装效果。 证据：`README.md` Claim：`clm_0019` 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：`agent-templates/packs/bio-informatics-pack/skills/bio-informatics-analyst/SKILL.md`, `agent-templates/packs/data-visualization-pack/skills/data-visualization-analyst/SKILL.md`, `agent-templates/packs/finance-analysis-pack/skills/finance-analysis-analyst/SKILL.md`, `agent-templates/packs/legal-research-pack/skills/legal-research-analyst/SKILL.md` 等 Claim：`clm_0020` supported 0.86 - **能力存在：Tandem Engine Runtime**（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：`engine/README.md`, `packages/tandem-engine/README.md`, `README.md`, `README.md` Claim：`clm_0001` supported 0.86, `clm_0003` supported 0.86, `clm_0006` supported 0.86, `clm_0009` supported 0.86 - **能力存在：Session and Run Management**（supported）：可以相信项目包含这类能力线索；是否适合你的具体任务仍要试用或安装后验证。 证据：`packages/tandem-client-ts/README.md`, `packages/tandem-client-py/README.md`, `packages/tandem-control-panel/README.md`, `docs/ENGINE_COMMUNICATION.md` Claim：`clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0005` supported 0.86, `clm_0006` supported 0.86 - **能力存在：Tool Execution with Approval Gates**（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`, `packages/tandem-tui/README.md`, `packages/tandem-control-panel/README.md`, `docs/ENGINE_CLI.md` 等 Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0006` supported 0.86 - **能力存在：MCP (Model Context Protocol) Integration**（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：`docs/MCP_IMPROVEMENTS.md`, `crates/tandem-server/resources/mcp-catalog/README.md`, `crates/tandem-server/resources/mcp-catalog/README.md`, `.agents/README.md` Claim：`clm_0004` supported 0.86 ### 现在还不能相信 - **研究结论、引用和实验结果不能在安装前相信。**（unverified）：研究 Skill 可以组织问题和路径，但不能替代真实资料检索、论文核验和实验复现。 - **是否适合你的具体研究领域不能直接相信。**（unverified）：Skill 覆盖很多研究主题，不代表对你的领域、资料要求和可信度标准足够。 - **真实输出质量不能在安装前相信。**（unverified）：Prompt Preview 只能展示引导方式，不能证明真实项目中的结果质量。 - **宿主 AI 版本兼容性不能在安装前相信。**（unverified）：Claude、Cursor、Codex、Gemini 等宿主加载规则和版本差异必须在真实环境验证。 - **不会污染现有宿主 AI 行为，不能直接相信。**（inferred）：Skill、plugin、AGENTS/CLAUDE/GEMINI 指令可能改变宿主 AI 的默认行为。 证据：`AGENTS.md`, `agent-templates/packs/bio-informatics-pack/skills/bio-informatics-analyst/SKILL.md`, `agent-templates/packs/data-visualization-pack/skills/data-visualization-analyst/SKILL.md`, `agent-templates/packs/finance-analysis-pack/skills/finance-analysis-analyst/SKILL.md` 等 - **可安全回滚不能默认相信。**（unverified）：除非项目明确提供卸载和恢复说明，否则必须先在隔离环境验证。 - **真实安装后是否与用户当前宿主 AI 版本兼容？**（unverified）：兼容性只能通过实际宿主环境验证。 - **项目输出质量是否满足用户具体任务？**（unverified）：安装前预览只能展示流程和边界，不能替代真实评测。 ### 继续会触碰什么 - **研究判断**：问题拆解、资料路径、实验路径、结论结构和可信度判断。 原因：研究型 Skill 可能让输出看起来更专业，但不能替代真实证据核验。 - **命令执行**：包管理器、网络下载、本地插件目录、项目配置或用户主目录。 原因：运行第一条命令就可能产生环境改动；必须先判断是否值得跑。 证据：`README.md`, `guide/src/content/docs/agent-workflow-mission-quickstart.md`, `packages/create-tandem-panel/README.md`, `packages/tandem-client-py/README.md` 等 - **宿主 AI 配置**：Claude/Codex/Cursor/Gemini/OpenCode 等宿主的 plugin、Skill 或规则加载配置。 原因：宿主配置会改变 AI 后续工作方式，可能和用户已有规则冲突。 证据：`AGENTS.md`, `agent-templates/packs/bio-informatics-pack/skills/bio-informatics-analyst/SKILL.md`, `agent-templates/packs/data-visualization-pack/skills/data-visualization-analyst/SKILL.md`, `agent-templates/packs/finance-analysis-pack/skills/finance-analysis-analyst/SKILL.md` 等 - **本地环境或项目文件**：安装结果、插件缓存、项目配置或本地依赖目录。 原因：安装前无法证明写入范围和回滚方式，需要隔离验证。 证据：`.agents/README.md`, `README.md`, `crates/tandem-plan-compiler/README.md`, `crates/tandem-server/resources/mcp-catalog/README.md` 等 - **环境变量 / API Key**：项目入口文档明确出现 API key、token、secret 或账号凭证配置。 原因：如果真实安装需要凭证，应先使用测试凭证并经过权限/合规判断。 证据：`docs/MCP_IMPROVEMENTS.md`, `docs/dev/EVAL_FRAMEWORK.md`, `guide/src/content/docs/autonomous-coding-agents-github-projects.md`, `guide/src/content/docs/choosing-providers-and-models-for-agents.md` 等 - **宿主 AI 上下文**：AI Context Pack、Prompt Preview、Skill 路由、风险规则和项目事实。 原因：导入上下文会影响宿主 AI 后续判断，必须避免把未验证项包装成事实。 ### 最小安全下一步 - **先跑 Prompt Preview**：先验证它能否正确界定研究问题和证据边界，不要先相信研究输出。（适用：任何项目都适用，尤其是输出质量未知时。） - **只在隔离目录或测试账号试装**：避免安装命令污染主力宿主 AI、真实项目或用户主目录。（适用：存在命令执行、插件配置或本地写入线索时。） - **先备份宿主 AI 配置**：Skill、plugin、规则文件可能改变 Claude/Cursor/Codex 的默认行为。（适用：存在插件 manifest、Skill 或宿主规则入口时。） - **不要使用真实生产凭证**：环境变量/API key 一旦进入宿主或工具链，可能产生账号和合规风险。（适用：出现 API、TOKEN、KEY、SECRET 等环境线索时。） - **安装后只验证一个最小任务**：先验证加载、兼容、输出质量和回滚，再决定是否深用。（适用：准备从试用进入真实工作流时。） ### 退出方式 - **保留安装前状态**：记录原始宿主配置和项目状态，后续才能判断是否可恢复。 - **准备移除宿主 plugin / Skill / 规则入口**：如果试装后行为异常，可以把宿主 AI 恢复到试装前状态。 - **保留资料和结论核验清单**：如果后续发现引用或实验路径不可靠，可以回到证据边界阶段重新校验。 - **记录安装命令和写入路径**：没有明确卸载说明时，至少要知道哪些目录或配置需要手动清理。 - **准备撤销测试 API key 或 token**：测试凭证泄露或误用时，可以快速止损。 - **如果没有回滚路径，不进入主力环境**：不可回滚是继续前阻断项，不应靠信任或运气继续。 ## 哪些只能预览 - 解释项目适合谁和能做什么 - 基于项目文档演示典型对话流程 - 帮助用户判断是否值得安装或继续研究 ## 哪些必须安装后验证 - 真实安装 Skill、插件或 CLI - 执行脚本、修改本地文件或访问外部服务 - 验证真实输出质量、性能和兼容性 ## 边界与风险判断卡 - **把安装前预览误认为真实运行**：用户可能高估项目已经完成的配置、权限和兼容性验证。 处理方式：明确区分 prompt_preview_can_do 与 runtime_required。 Claim：`clm_0030` 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`, `guide/src/content/docs/agent-workflow-mission-quickstart.md`, `packages/create-tandem-panel/README.md`, `packages/tandem-client-py/README.md` 等 Claim：`clm_0031` supported 0.86 - **风险**： 处理方式： - **风险**： 处理方式： - **风险**： 处理方式： - **风险**： 处理方式： - **风险**： 处理方式： - **风险**： 处理方式： - **待确认**：真实安装后是否与用户当前宿主 AI 版本兼容？。原因：兼容性只能通过实际宿主环境验证。 - **待确认**：项目输出质量是否满足用户具体任务？。原因：安装前预览只能展示流程和边界，不能替代真实评测。 - **待确认**：安装命令是否需要网络、权限或全局写入？。原因：这影响企业环境和个人环境的安装风险。 ## 开工前工作上下文 ### 加载顺序 - 先读取 how_to_use.host_ai_instruction，建立安装前判断资产的边界。 - 读取 claim_graph_summary，确认事实来自 Claim/Evidence Graph，而不是 Human Wiki 叙事。 - 再读取 intended_users、capabilities 和 quick_start_candidates，判断用户是否匹配。 - 需要执行具体任务时，优先查 role_skill_index，再查 evidence_index。 - 遇到真实安装、文件修改、网络访问、性能或兼容性问题时，转入 risk_card 和 boundaries.runtime_required。 ### 任务路由 - **Tandem Engine Runtime**：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：`engine/README.md`, `packages/tandem-engine/README.md`, `README.md`, `README.md` Claim：`clm_0001` supported 0.86, `clm_0003` supported 0.86, `clm_0006` supported 0.86, `clm_0009` supported 0.86 等 - **Session and Run Management**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`packages/tandem-client-ts/README.md`, `packages/tandem-client-py/README.md`, `packages/tandem-control-panel/README.md`, `docs/ENGINE_COMMUNICATION.md` Claim：`clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0005` supported 0.86, `clm_0006` supported 0.86 等 - **Tool Execution with Approval Gates**：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`, `packages/tandem-tui/README.md`, `packages/tandem-control-panel/README.md`, `docs/ENGINE_CLI.md` 等 Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0006` supported 0.86 等 - **MCP (Model Context Protocol) Integration**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`docs/MCP_IMPROVEMENTS.md`, `crates/tandem-server/resources/mcp-catalog/README.md`, `crates/tandem-server/resources/mcp-catalog/README.md`, `.agents/README.md` Claim：`clm_0004` supported 0.86 - **Browser Automation Tools**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`packages/tandem-client-py/README.md`, `packages/tandem-client-py/README.md`, `packages/tandem-enterprise/README.md` Claim：`clm_0002` supported 0.86, `clm_0005` supported 0.86, `clm_0010` supported 0.86, `clm_0011` supported 0.86 等 - **Permissioned Memory Management**：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：`engine/README.md`, `engine/README.md`, `README.md`, `packages/tandem-control-panel/README.md` Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0006` supported 0.86 等 - **Scheduled and On-Demand Routines**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`packages/tandem-client-ts/README.md`, `packages/tandem-client-ts/README.md`, `packages/tandem-control-panel/README.md` Claim：`clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0006` supported 0.86, `clm_0007` supported 0.86 等 - **Workflow Plan Compiler**：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：`crates/tandem-plan-compiler/README.md`, `crates/tandem-plan-compiler/README.md`, `packages/tandem-client-ts/README.md` Claim：`clm_0002` supported 0.86, `clm_0007` supported 0.86, `clm_0008` supported 0.86, `clm_0010` supported 0.86 - **Multi-Client Entry Points**：先基于 role_skill_index / evidence_index 帮用户挑选可用角色、Skill 或工作流。 边界：可做安装前 Prompt 体验。 证据：`README.md`, `apps/tandem-desktop/package.json`, `crates/tandem-tui/README.md`, `packages/tandem-control-panel/README.md` 等 Claim：`clm_0001` supported 0.86, `clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0006` supported 0.86 等 - **Python and TypeScript SDKs**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`packages/tandem-client-py/README.md`, `packages/tandem-client-py/README.md`, `packages/tandem-client-ts/README.md`, `packages/tandem-client-ts/package.json` Claim：`clm_0002` supported 0.86, `clm_0005` supported 0.86, `clm_0007` supported 0.86, `clm_0008` supported 0.86 等 - **Multi-Channel Integration**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`packages/tandem-control-panel/README.md`, `docs/CHANNELS_LIFECYCLE_AND_DIAGNOSTICS.md`, `packages/tandem-client-py/README.md` Claim：`clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0005` supported 0.86, `clm_0006` supported 0.86 等 - **Provider-Agnostic LLM Routing**：先基于 role_skill_index / evidence_index 帮用户挑选可用角色、Skill 或工作流。 边界：可做安装前 Prompt 体验。 证据：`README.md`, `packages/tandem-engine/README.md`, `engine/README.md` Claim：`clm_0001` supported 0.86, `clm_0003` supported 0.86, `clm_0006` supported 0.86, `clm_0009` supported 0.86 等 - **Enterprise Linux Binary Distribution**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`packages/tandem-enterprise/README.md`, `packages/tandem-enterprise/package.json` Claim：`clm_0005` supported 0.86, `clm_0013` supported 0.86 - **EU AI Act Compliance Documentation**：先基于 role_skill_index / evidence_index 帮用户挑选可用角色、Skill 或工作流。 边界：可做安装前 Prompt 体验。 证据：`docs/compliance/README.md`, `docs/EU_AI_ACT_COMPLIANCE.md` Claim：`clm_0014` supported 0.86 - **Concurrent Batch Processing**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`packages/tandem-engine/README.md`, `packages/tandem-engine/README.md` Claim：`clm_0001` supported 0.86, `clm_0012` supported 0.86, `clm_0015` supported 0.86 - **Web Research Tools**：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：`docs/ENGINE_CLI.md`, `docs/ENGINE_CLI.md`, `scripts/bench-js/README.md` Claim：`clm_0003` supported 0.86, `clm_0016` supported 0.86, `clm_0017` supported 0.86 - **Token Generation and API Authentication**：先基于 role_skill_index / evidence_index 帮用户挑选可用角色、Skill 或工作流。 边界：可做安装前 Prompt 体验。 证据：`packages/tandem-client-py/README.md`, `docs/ENGINE_CLI.md`, `docs/ENGINE_COMMUNICATION.md` Claim：`clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0005` supported 0.86, `clm_0010` supported 0.86 等 - **Swarm Orchestration with Live Flow Visualization**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`packages/tandem-control-panel/README.md`, `docs/ENGINE_RUNTIME.md` Claim：`clm_0002` supported 0.86, `clm_0003` supported 0.86, `clm_0006` supported 0.86, `clm_0007` supported 0.86 等 ### 上下文规模 - 文件总数：2330 - 重要文件覆盖：40/2330 - 证据索引条目：178 - 角色 / Skill 条目：121 ### 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 tandem 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 tandem 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 tandem 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 121 个角色 / Skill / 项目文档条目。 - **bio-informatics-analyst**（skill）：Coordinate bioinformatics data conversion, pipeline setup, and analysis reporting with reproducible workflow checkpoints. 激活提示：当用户任务与“bio-informatics-analyst”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：`agent-templates/packs/bio-informatics-pack/skills/bio-informatics-analyst/SKILL.md` - **data-visualization-analyst**（skill）：Guide data cleanup and chart design into publication-quality visual outputs with clear narrative framing. 激活提示：当用户任务与“data-visualization-analyst”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：`agent-templates/packs/data-visualization-pack/skills/data-visualization-analyst/SKILL.md` - **finance-analysis-analyst**（skill）：Drive repeatable finance variance analysis with clear assumptions, reconciliations, and executive summaries. 激活提示：当用户任务与“finance-analysis-analyst”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：`agent-templates/packs/finance-analysis-pack/skills/finance-analysis-analyst/SKILL.md` - **legal-research-analyst**（skill）：Perform structured legal document analysis, issue spotting, and memo drafting with clear risk framing. 激活提示：当用户任务与“legal-research-analyst”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：`agent-templates/packs/legal-research-pack/skills/legal-research-analyst/SKILL.md` - **micro-drama-script-studio**（skill）：Guide short-form episodic script production with consistent beats, character voice, and production-ready outputs. 激活提示：当用户任务与“micro-drama-script-studio”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：`agent-templates/packs/micro-drama-script-studio-pack/skills/micro-drama-script-studio/SKILL.md` - **research-synthesis-analyst**（skill）：Synthesize multi-source evidence into decision-ready research outputs with conflict tracking and citation hygiene. 激活提示：当用户任务与“research-synthesis-analyst”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：`agent-templates/packs/research-synthesis-pack/skills/research-synthesis-analyst/SKILL.md` - **security-playbook-builder**（skill）：Transform security context into prioritized controls, response runbooks, and compliance-ready documentation. 激活提示：当用户任务与“security-playbook-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：`agent-templates/packs/security-playbook-pack/skills/security-playbook-builder/SKILL.md` - **web-research-refresh**（skill）：Refresh stale content via web verification, source logging, and citation-backed updates. 激活提示：当用户任务与“web-research-refresh”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：`agent-templates/packs/web-research-refresh-pack/skills/web-research-refresh/SKILL.md` - **web-starter-audit**（skill）：Run a full UX, accessibility, and quality audit for starter web projects and produce prioritized remediation. 激活提示：当用户任务与“web-starter-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：`agent-templates/packs/web-starter-audit-pack/skills/web-starter-audit/SKILL.md` - **bio-informatics-analyst**（skill）：Coordinate bioinformatics data conversion, pipeline setup, and analysis reporting with reproducible workflow checkpoints. 激活提示：当用户任务与“bio-informatics-analyst”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：`apps/tandem-desktop/src-tauri/resources/packs/bio-informatics-pack/skills/bio-informatics-analyst/SKILL.md` - **data-visualization-analyst**（skill）：Guide data cleanup and chart design into publication-quality visual outputs with clear narrative framing. 激活提示：当用户任务与“data-visualization-analyst”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：`apps/tandem-desktop/src-tauri/resources/packs/data-visualization-pack/skills/data-visualization-analyst/SKILL.md` - **finance-analysis-analyst**（skill）：Drive repeatable finance variance analysis with clear assumptions, reconciliations, and executive summaries. 激活提示：当用户任务与“finance-analysis-analyst”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：`apps/tandem-desktop/src-tauri/resources/packs/finance-analysis-pack/skills/finance-analysis-analyst/SKILL.md` - **legal-research-analyst**（skill）：Perform structured legal document analysis, issue spotting, and memo drafting with clear risk framing. 激活提示：当用户任务与“legal-research-analyst”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：`apps/tandem-desktop/src-tauri/resources/packs/legal-research-pack/skills/legal-research-analyst/SKILL.md` - **micro-drama-script-studio**（skill）：Guide short-form episodic script production with consistent beats, character voice, and production-ready outputs. 激活提示：当用户任务与“micro-drama-script-studio”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：`apps/tandem-desktop/src-tauri/resources/packs/micro-drama-script-studio-pack/skills/micro-drama-script-studio/SKILL.md` - **research-synthesis-analyst**（skill）：Synthesize multi-source evidence into decision-ready research outputs with conflict tracking and citation hygiene. 激活提示：当用户任务与“research-synthesis-analyst”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：`apps/tandem-desktop/src-tauri/resources/packs/research-synthesis-pack/skills/research-synthesis-analyst/SKILL.md` - **security-playbook-builder**（skill）：Transform security context into prioritized controls, response runbooks, and compliance-ready documentation. 激活提示：当用户任务与“security-playbook-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：`apps/tandem-desktop/src-tauri/resources/packs/security-playbook-pack/skills/security-playbook-builder/SKILL.md` - **web-research-refresh**（skill）：Refresh stale content via web verification, source logging, and citation-backed updates. 激活提示：当用户任务与“web-research-refresh”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：`apps/tandem-desktop/src-tauri/resources/packs/web-research-refresh-pack/skills/web-research-refresh/SKILL.md` - **web-starter-audit**（skill）：Run a full UX, accessibility, and quality audit for starter web projects and produce prioritized remediation. 激活提示：当用户任务与“web-starter-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：`apps/tandem-desktop/src-tauri/resources/packs/web-starter-audit-pack/skills/web-starter-audit/SKILL.md` - **algorithmic-art**（skill）：Creating algorithmic art using p5.js with seeded randomness and interactive parameter exploration. Use this when users request creating art using code, generative art, algorithmic art, flow fields, or particle systems. Create original algorithmic art rather than copying existing artists' work to avoid copyright violations. 激活提示：当用户任务与“algorithmic-art”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/algorithmic-art/SKILL.md` - **bio-instrument-data**（skill）：Convert laboratory instrument output files PDF, CSV, Excel, TXT to Allotrope Simple Model ASM JSON format or flattened 2D CSV. Use this skill when scientists need to standardize instrument data for LIMS systems, data lakes, or downstream analysis. Supports auto-detection of instrument types. Outputs include full ASM JSON, flattened CSV for easy import, and exportable Python code for data engineers. Common triggers i… 激活提示：当用户任务与“bio-instrument-data”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/bio-instrument-data/SKILL.md` - **bio-nextflow-manager**（skill）：Run nf-core bioinformatics pipelines rnaseq, sarek, atacseq on sequencing data. Use when analyzing RNA-seq, WGS/WES, or ATAC-seq data—either local FASTQs or public datasets from GEO/SRA. Triggers on nf-core, Nextflow, FASTQ analysis, variant calling, gene expression, differential expression, GEO reanalysis, GSE/GSM/SRR accessions, or samplesheet creation. 激活提示：当用户任务与“bio-nextflow-manager”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/bio-nextflow-manager/SKILL.md` - **bio-research-strategy**（skill）：Strategic scientific problem selection, project ideation, and troubleshooting based on the Fischbach & Walsh framework. 激活提示：当用户任务与“bio-research-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：`apps/tandem-desktop/src-tauri/resources/skill-templates/bio-research-strategy/SKILL.md` - **bio-single-cell**（skill）：Deep learning for single-cell analysis using scvi-tools and scverse ecosystem. This skill should be used when users need 1 data integration and batch correction with scVI/scANVI, 2 ATAC-seq analysis with PeakVI, 3 CITE-seq multi-modal analysis with totalVI, 4 multiome RNA+ATAC analysis with MultiVI, 5 spatial transcriptomics deconvolution with DestVI, 6 label transfer and reference mapping, 7 RNA velocity with veloV… 激活提示：当用户任务与“bio-single-cell”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/bio-single-cell/SKILL.md` - **bio-strategy**（skill）：A conversational framework for systematic scientific problem selection, project ideation, troubleshooting, and strategic decision making. 激活提示：当用户任务与“bio-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：`apps/tandem-desktop/src-tauri/resources/skill-templates/bio-strategy/SKILL.md` - **brainstorming**（skill）：You MUST use this before any creative work features, products, content, strategy, systems, or behavior changes . Start by classifying what we’re brainstorming, then run thorough one-question-at-a-time discovery, propose 2–3 approaches, and converge on a validated plan/spec. 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：`apps/tandem-desktop/src-tauri/resources/skill-templates/brainstorming/SKILL.md` - **canvas-design**（skill）：Create beautiful visual art in .png and .pdf documents using design philosophy. You should use this skill when the user asks to create a poster, piece of art, design, or other static piece. Create original visual designs, never copying existing artists' work to avoid copyright violations. 激活提示：当用户任务与“canvas-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：`apps/tandem-desktop/src-tauri/resources/skill-templates/canvas-design/SKILL.md` - **competitor-alternatives**（skill）：Plan and draft competitor comparison and alternative-page strategies for SEO and sales evaluation use cases. Use for " competitor alternatives", "you vs competitor", and comparison-page content architecture. Produce deterministic outputs under scripts/marketing/ / with verified claim handling. 激活提示：当用户任务与“competitor-alternatives”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/competitor-alternatives/SKILL.md` - **competitor-price-tracker**（skill）：Monitor competitor product pages and alert when prices change. 激活提示：当用户任务与“competitor-price-tracker”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/competitor-price-tracker/SKILL.md` - **content-strategy**（skill）：Create a practical content strategy that connects searchable and shareable content to business goals. Use when planning topics, clusters, editorial calendar, repurposing, and channel priorities. Produce deterministic file outputs under scripts/marketing/ / and require web research when available. 激活提示：当用户任务与“content-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：`apps/tandem-desktop/src-tauri/resources/skill-templates/content-strategy/SKILL.md` - **copy-editing**（skill）：Improve existing marketing copy through structured editing passes for clarity, voice, proof, specificity, and conversion impact. Use after drafting copy for pages, emails, and campaign assets. Produce deterministic review artifacts under scripts/marketing/ /. 激活提示：当用户任务与“copy-editing”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/copy-editing/SKILL.md` - **copywriting**（skill）：Draft conversion-focused marketing copy for landing pages, homepages, pricing pages, feature pages, and campaign assets. Use when creating first drafts from positioning and audience context. Output file-first drafts under scripts/marketing/ / with explicit proof and CTA requirements. 激活提示：当用户任务与“copywriting”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/copywriting/SKILL.md` - **cowork-mcp-config-assistant**（skill）：Configure Model Context Protocol MCP servers and customize Tandem skills for your organization. 激活提示：当用户任务与“cowork-mcp-config-assistant”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/cowork-mcp-config-assistant/SKILL.md` - **data-analyze**（skill）：Answer data questions -- from quick lookups to full analyses 激活提示：当用户任务与“data-analyze”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/data-analyze/SKILL.md` - **data-build-dashboard**（skill）：Build an interactive HTML dashboard with charts, filters, and tables 激活提示：当用户任务与“data-build-dashboard”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/data-build-dashboard/SKILL.md` - **data-create-viz**（skill）：Create publication-quality visualizations with Python 激活提示：当用户任务与“data-create-viz”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/data-create-viz/SKILL.md` - **data-explore-data**（skill）：Profile and explore a dataset to understand its shape, quality, and patterns 激活提示：当用户任务与“data-explore-data”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/data-explore-data/SKILL.md` - **data-validate**（skill）：QA an analysis before sharing -- methodology, accuracy, and bias checks 激活提示：当用户任务与“data-validate”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/data-validate/SKILL.md` - **data-write-query**（skill）：Write optimized SQL for your dialect with best practices 激活提示：当用户任务与“data-write-query”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/data-write-query/SKILL.md` - **dev-agent**（skill）：Analyze repositories, implement code changes, run tests, and optionally commit. 激活提示：当用户任务与“dev-agent”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/dev-agent/SKILL.md` - **development-estimation**（skill）：Use when estimating time, effort, cost, or complexity for features, projects, refactors, and bug backlogs. Produces defensible estimates via triage, decomposition, risk handling, and confidence intervals, with clear assumptions and validation steps. 激活提示：当用户任务与“development-estimation”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/development-estimation/SKILL.md` - **email-digest**（skill）：Summarize important unread emails each morning. 激活提示：当用户任务与“email-digest”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/email-digest/SKILL.md` - **email-sequence**（skill）：Design and draft lifecycle email sequences such as welcome, onboarding, nurture, re-engagement, and upgrade flows. Use when defining triggers, sequence maps, email copy, and performance metrics. Produce file-first outputs under scripts/marketing/ / and align with product marketing context. 激活提示：当用户任务与“email-sequence”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/email-sequence/SKILL.md` - **enterprise-knowledge-synthesis**（skill）：Combines search results from multiple sources into coherent, deduplicated answers with source attribution. 激活提示：当用户任务与“enterprise-knowledge-synthesis”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/enterprise-knowledge-synthesis/SKILL.md` - **enterprise-search-knowledge**（skill）：Combines search results from multiple sources into coherent, deduplicated answers with source attribution. Handles confidence scoring based on freshness and authority, and summarizes large result sets effectively. 激活提示：当用户任务与“enterprise-search-knowledge”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/enterprise-search-knowledge/SKILL.md` - **enterprise-search-source**（skill）：Manages connected MCP sources for Tandem search. Detects available sources, guides users to connect new ones, handles source priority ordering, and manages rate limiting awareness. 激活提示：当用户任务与“enterprise-search-source”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/enterprise-search-source/SKILL.md` - **enterprise-search-strategy**（skill）：Query decomposition and multi-source search orchestration. Breaks natural language questions into targeted searches per source, translates queries into source-specific syntax, ranks results by relevance, and handles ambiguity and fallback strategies. 激活提示：当用户任务与“enterprise-search-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：`apps/tandem-desktop/src-tauri/resources/skill-templates/enterprise-search-strategy/SKILL.md` - **enterprise-source-management**（skill）：Manages connected MCP sources for enterprise search. Detects available sources and handles source priority ordering. 激活提示：当用户任务与“enterprise-source-management”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/enterprise-source-management/SKILL.md` - **finance-income-statement**（skill）：Generate an income statement with period-over-period comparison and variance analysis 激活提示：当用户任务与“finance-income-statement”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/finance-income-statement/SKILL.md` - **finance-journal-entry**（skill）：Prepare journal entries with proper debits, credits, and supporting detail 激活提示：当用户任务与“finance-journal-entry”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/finance-journal-entry/SKILL.md` - **finance-reconciliation**（skill）：Reconcile GL balances to subledger, bank, or third-party balances 激活提示：当用户任务与“finance-reconciliation”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/finance-reconciliation/SKILL.md` - **finance-sox-testing**（skill）：Generate SOX sample selections, testing workpapers, and control assessments 激活提示：当用户任务与“finance-sox-testing”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/finance-sox-testing/SKILL.md` - **finance-variance-analysis**（skill）：Decompose variances into drivers with narrative explanations and waterfall analysis 激活提示：当用户任务与“finance-variance-analysis”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`apps/tandem-desktop/src-tauri/resources/skill-templates/finance-variance-analysis/SKILL.md` - **frontend-design**（skill）：Create distinctive, production-grade frontend interfaces with high design quality. Use this skill when the user asks to build web components, pages, artifacts, posters, or applications examples include websites, landing pages, dashboards, React components, HTML/CSS layouts, or when styling/beautifying any web UI . Generates creative, polished code and UI design that avoids generic AI aesthetics. 激活提示：当用户任务与“frontend-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：`apps/tandem-desktop/src-tauri/resources/skill-templates/frontend-design/SKILL.md` - **github-issue-reporter**（skill）：Convert incidents into structured GitHub issues. 激活提示：当用户任务与“github-issue-reporter”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/github-issue-reporter/SKILL.md` - **inbox-triage**（skill）：Categorize messages and propose replies and next steps. 激活提示：当用户任务与“inbox-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：`apps/tandem-desktop/src-tauri/resources/skill-templates/inbox-triage/SKILL.md` - **intent-to-mission-blueprint**（skill）：Convert human operational intent into a Tandem mission blueprint with phased workstreams, explicit handoffs, reusable outputs, and recurrence-aware setup. 激活提示：当用户任务与“intent-to-mission-blueprint”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/intent-to-mission-blueprint/SKILL.md` - **internal-comms**（skill）：A set of resources to help me write all kinds of internal communications, using the formats that my company likes to use. Use this skill whenever asked to write internal communications status reports, leadership updates, 3P updates, company newsletters, FAQs, incident reports, project updates, etc. . 激活提示：当用户任务与“internal-comms”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/internal-comms/SKILL.md` - **launch-strategy**（skill）：Plan phased product and feature launches across owned, rented, and borrowed channels. Use when coordinating prelaunch, launch-day, and postlaunch execution with clear owners and assets. Produce deterministic runbooks under scripts/marketing/ /. 激活提示：当用户任务与“launch-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：`apps/tandem-desktop/src-tauri/resources/skill-templates/launch-strategy/SKILL.md` - **canned-responses**（skill）：Generate templated responses for common legal inquiries and identify when situations require individualized attention. Use when responding to routine legal questions — data subject requests, vendor inquiries, NDA requests, discovery holds — or when managing response templates. 激活提示：当用户任务与“canned-responses”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/legal-canned-responses/SKILL.md` - **compliance**（skill）：Navigate privacy regulations GDPR, CCPA , review DPAs, and handle data subject requests. Use when reviewing data processing agreements, responding to data subject access or deletion requests, assessing cross-border data transfer requirements, or evaluating privacy compliance. 激活提示：当用户任务与“compliance”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/legal-compliance/SKILL.md` - **contract-review**（skill）：Review contracts against your organization's negotiation playbook, flagging deviations and generating redline suggestions. Use when reviewing vendor contracts, customer agreements, or any commercial agreement where you need clause-by-clause analysis against standard positions. 激活提示：当用户任务与“contract-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：`apps/tandem-desktop/src-tauri/resources/skill-templates/legal-contract-review/SKILL.md` - **meeting-briefing**（skill）：Prepare structured briefings for meetings with legal relevance and track resulting action items. Use when preparing for contract negotiations, board meetings, compliance reviews, or any meeting where legal context, background research, or action tracking is needed. 激活提示：当用户任务与“meeting-briefing”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/legal-meeting-briefing/SKILL.md` - **nda-triage**（skill）：Screen incoming NDAs and classify them as GREEN standard , YELLOW needs review , or RED significant issues . Use when a new NDA comes in from sales or business development, when assessing NDA risk level, or when deciding whether an NDA needs full counsel review. 激活提示：当用户任务与“nda-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：`apps/tandem-desktop/src-tauri/resources/skill-templates/legal-nda-triage/SKILL.md` - **legal-risk-assessment**（skill）：Assess and classify legal risks using a severity-by-likelihood framework with escalation criteria. Use when evaluating contract risk, assessing deal exposure, classifying issues by severity, or determining whether a matter needs senior counsel or outside legal review. 激活提示：当用户任务与“legal-risk-assessment”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/legal-risk-assessment/SKILL.md` - **marketing-brand-voice**（skill）：Legacy fallback for brand governance and style policy work. Prefer copywriting and copy-editing for operational production tasks. Use when documenting or auditing voice rules, not as the default drafting workflow. 激活提示：当用户任务与“marketing-brand-voice”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/marketing-brand-voice/SKILL.md` - **marketing-campaign-planning**（skill）：Legacy fallback for high-level campaign PM structure. Prefer launch-strategy for launch execution and canonical Core 9 skills for channel work. Use this template when a generic campaign framework is explicitly needed. 激活提示：当用户任务与“marketing-campaign-planning”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/marketing-campaign-planning/SKILL.md` - **marketing-competitive-analysis**（skill）：Legacy fallback for broad competitor intelligence. Prefer competitor-alternatives for conversion-focused comparison and alternative page execution. Use this template when a wide market landscape brief is explicitly requested. 激活提示：当用户任务与“marketing-competitive-analysis”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`apps/tandem-desktop/src-tauri/resources/skill-templates/marketing-competitive-analysis/SKILL.md` - **marketing-content-creation**（skill）：Legacy fallback for broad marketing content requests. Prefer canonical skills for routine execution: copywriting, copy-editing, social-content, and email-sequence. Use only when a single broad multi-channel fallback is explicitly requested. 激活提示：当用户任务与“marketing-content-creation”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/marketing-content-creation/SKILL.md` - **marketing-performance-analytics**（skill）：Analyze marketing performance with key metrics, trend analysis, and optimization recommendations. Use when building performance reports, reviewing campaign results, analyzing channel metrics email, social, paid, SEO , or identifying what's working and what needs improvement. 激活提示：当用户任务与“marketing-performance-analytics”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/marketing-performance-analytics/SKILL.md` - **marketing-research-posting-plan**（skill）：Legacy fallback for heavy research-plus-posting file packs. Prefer content-strategy, social-content, and seo-audit for day-to-day planning and execution. Use this only when a full research dossier and posting package is explicitly required. 激活提示：当用户任务与“marketing-research-posting-plan”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/marketing-research-posting-plan/SKILL.md` - **meeting-notes**（skill）：Analyze meeting transcripts and notes from your folder to generate structured minutes, decisions, and action items. 激活提示：当用户任务与“meeting-notes”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/meeting-notes/SKILL.md` - **mission-blueprint-compiler**（skill）：Design Tandem mission blueprints with scoped workstreams, explicit handoffs, review gates, and prompts that compile into reliable staged missions. 激活提示：当用户任务与“mission-blueprint-compiler”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/mission-blueprint-compiler/SKILL.md` - **mode-builder**（skill）：Create Tandem custom modes through guided questions, then output one valid mode JSON object for preview and apply. 激活提示：当用户任务与“mode-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：`apps/tandem-desktop/src-tauri/resources/skill-templates/mode-builder/SKILL.md` - **news-digest**（skill）：Gather and summarize topic-specific news. 激活提示：当用户任务与“news-digest”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/news-digest/SKILL.md` - **notion-weekly-report**（skill）：Produce weekly progress summaries from Notion sources. 激活提示：当用户任务与“notion-weekly-report”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/notion-weekly-report/SKILL.md` - **product-competitive-analysis**（skill）：Analyze competitors with feature comparison matrices, positioning analysis, and strategic implications. Use when researching a competitor, comparing product capabilities, assessing competitive positioning, or preparing a competitive brief for product strategy. 激活提示：当用户任务与“product-competitive-analysis”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`apps/tandem-desktop/src-tauri/resources/skill-templates/product-competitive-analysis/SKILL.md` - **product-feature-spec**（skill）：Write structured product requirements documents PRDs with problem statements, user stories, requirements, and success metrics. Use when speccing a new feature, writing a PRD, defining acceptance criteria, prioritizing requirements, or documenting product decisions. 激活提示：当用户任务与“product-feature-spec”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/product-feature-spec/SKILL.md` - **product-marketing-context**（skill）：Build and maintain the shared product marketing context used by other marketing skills. Use when setting positioning, ICP, messaging pillars, voice rules, proof points, or when other marketing tasks need foundational context. Produce file-first outputs under scripts/marketing/ / and update scripts/marketing/ shared/product-marketing-context.md. 激活提示：当用户任务与“product-marketing-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：`apps/tandem-desktop/src-tauri/resources/skill-templates/product-marketing-context/SKILL.md` - **product-metrics**（skill）：Define, track, and analyze product metrics with frameworks for goal setting and dashboard design. Use when setting up OKRs, building metrics dashboards, running weekly metrics reviews, identifying trends, or choosing the right metrics for a product area. 激活提示：当用户任务与“product-metrics”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/product-metrics/SKILL.md` - **product-roadmap**（skill）：Plan and prioritize product roadmaps using frameworks like RICE, MoSCoW, and ICE. Use when creating a roadmap, reprioritizing features, mapping dependencies, choosing between Now/Next/Later or quarterly formats, or presenting roadmap tradeoffs to stakeholders. 激活提示：当用户任务与“product-roadmap”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：`apps/tandem-desktop/src-tauri/resources/skill-templates/product-roadmap/SKILL.md` - 其余 41 个条目见 `AI_CONTEXT_PACK.json`。 ## 证据索引 - 共索引 178 条证据。 - **Tandem Documentation**（documentation）：This folder contains public technical references. 证据：`docs/README.md` - **Tandem Compliance Starter Pack**（documentation）：This folder contains public starter documents for teams evaluating Tandem in regulated or security-sensitive AI workflows. The documents are intended for CISOs, compliance teams, platform engineers, and deployers who need a practical view of what Tandem supports today, what remains the deployer's responsibility, and what Tandem plans to add. 证据：`docs/compliance/README.md` - **.agents Guide**（documentation）：Local agent instructions for this repo live here. 证据：`.agents/README.md` - **Why Tandem Exists**（documentation）：Interested in Tandem Hosted? Join the waitlist 证据：`README.md` - **tandem-ai tandem-engine**（documentation）：tandem-ai is the Rust crate that builds the tandem-engine binary. It runs the headless Tandem runtime HTTP + SSE APIs, tools, sessions, orchestration/agent workflows . 证据：`engine/README.md` - **Our Site - Web Project**（documentation）：This is the source code for Our Site e-commerce platform. 证据：`agent-templates/packs/web-starter-audit-pack/src/README.md` - **Readme**（documentation）：This directory is intentionally tracked so Tauri resource resolution does not fail in CI when no bundled engine binary is present. 证据：`apps/tandem-desktop/src-tauri/resources/binaries/README.md` - **Our Site - Web Project**（documentation）：This is the source code for Our Site e-commerce platform. 证据：`apps/tandem-desktop/src-tauri/resources/packs/web-starter-audit-pack/src/README.md` - **tandem-plan-compiler**（documentation）：Mission and plan compiler boundary for Tandem. 证据：`crates/tandem-plan-compiler/README.md` - **MCP Catalog Generated**（documentation）：- Sources: - Anthropic MCP registry https://api.anthropic.com/mcp-registry/v0/servers - Curated additions curated-mcp-overrides - Generated at: 2026-03-02T23:03:19.833Z - Version: latest - Visibility: commercial - Servers: 162 证据：`crates/tandem-server/resources/mcp-catalog/README.md` - **tandem-tui**（documentation）：tandem-tui is the Rust crate for the terminal client binary. It connects to tandem-engine and provides a keyboard-first chat + agent workflow UI in the terminal. 证据：`crates/tandem-tui/README.md` - **create-tandem-panel**（documentation）：Scaffold a fully editable Tandem control panel app. 证据：`packages/create-tandem-panel/README.md` - **PROJECT NAME**（documentation）：Editable Tandem control panel app scaffold. 证据：`packages/create-tandem-panel/template/README.md` - **tandem-client**（documentation）：Python client for the Tandem https://tandem.ac/ autonomous agent engine HTTP + SSE API. 证据：`packages/tandem-client-py/README.md` - **@frumu/tandem-client**（documentation）：TypeScript / Node.js client for the Tandem https://tandem.ac/ autonomous agent engine HTTP + SSE API. 证据：`packages/tandem-client-ts/README.md` - **Tandem Control Panel**（documentation）：Full web control center for Tandem Engine non-desktop entry point . 证据：`packages/tandem-control-panel/README.md` - **Tandem Master CLI and Engine Wrapper**（documentation）：Tandem Master CLI and Engine Wrapper 证据：`packages/tandem-engine/README.md` - **@frumu/tandem-enterprise**（documentation）：Hosted enterprise Tandem engine binary distribution for Linux x64. 证据：`packages/tandem-enterprise/README.md` - **Tandem TUI npm Wrapper**（documentation）：Prebuilt npm distribution of Tandem TUI for macOS, Linux, and Windows. Installing this package gives you the tandem-tui terminal client binary without compiling Rust locally. 证据：`packages/tandem-tui/README.md` - **JS Benchmark for WebFetch**（documentation）：This directory contains a Node.js benchmark script to compare against the Rust webfetch tool. 证据：`scripts/bench-js/README.md` - **esaxx-rs**（documentation）：! https://github.com/Narsil/esaxx-rs/workflows/build/badge.svg 证据：`third_party/esaxx-rs/README.md` - **Package**（package_manifest）：{ "name": "tandem-guide", "private": true, "type": "module", "scripts": { "prebuild": "node scripts/write-docs-build-metadata.mjs", "dev": "astro dev", "build": "astro build", "check": "astro check", "preview": "astro preview", "astro": "astro" }, "dependencies": { "@astrojs/starlight": "^0.32.2", "astro": "^5.5.0", "astro-mermaid": "^1.3.1", "mermaid": "^11.12.3" }, "devDependencies": { "@astrojs/check": "^0.9.4", "typescript": "^5.9.3" } } 证据：`guide/package.json` - **Package**（package_manifest）：{ "name": "tandem", "private": true, "version": "0.5.11", "type": "module", "scripts": { "desktop:dev": "pnpm -C apps/tandem-desktop tauri dev", "desktop:build": "pnpm -C apps/tandem-desktop tauri build", "desktop:preview": "pnpm -C apps/tandem-desktop preview", "desktop:install": "pnpm -C apps/tandem-desktop install", "engine:stop:windows": "powershell -NoProfile -ExecutionPolicy Bypass -Command \"Get-Process Where-Object { $ .ProcessName -in @ 'tandem-engine','tandem','tandem-tui' } Stop-Process -Force -ErrorAction SilentlyContinue\"", "engine:build": "cargo build -p tandem-ai", "agents:catalog:refresh": "node scripts/generate-agent-catalog.mjs", "agents:catalog:test": "node --test script… 证据：`package.json` - **Package**（package_manifest）：{ "name": "@frumu/tandem-desktop", "private": true, "version": "0.5.11", "type": "module", "scripts": { "dev": "vite", "build": "tsc && vite build", "preview": "vite preview", "tauri": "tauri", "tauri:dev:clean": "powershell -NoProfile -ExecutionPolicy Bypass -Command \"Get-Process Where-Object { $ .ProcessName -in @ 'tandem','tandem-engine','node' } Stop-Process -Force -ErrorAction SilentlyContinue; pnpm tauri dev\"", "test:blackboard": "rm -rf .tmp/ts-tests && tsc -p tsconfig.tests.json && node --test .tmp/ts-tests/ .test.js", "lint": "eslint src", "lint:fix": "eslint src --fix", "format": "prettier --write \"src/ / .{ts,tsx}\"", "format:check": "prettier --check \"src/ / .{ts,tsx}\"" },… 证据：`apps/tandem-desktop/package.json` - **Package**（package_manifest）：{ "name": "create-tandem-panel", "version": "0.5.11", "description": "Scaffold an editable Tandem authority-layer control panel app", "type": "module", "bin": { "create-tandem-panel": "index.js" }, "files": "index.js", "template", "README.md" , "scripts": { "test": "node --test tests/ .test.mjs" }, "repository": { "type": "git", "url": "git+https://github.com/frumu-ai/tandem.git", "directory": "packages/create-tandem-panel" }, "homepage": "https://tandem.ac", "author": "Frumu Ltd", "license": "MIT OR Apache-2.0", "publishConfig": { "access": "public" }, "engines": { "node": " =18" } } 证据：`packages/create-tandem-panel/package.json` - **Package**（package_manifest）：{ "name": " PROJECT NAME ", "private": true, "version": "0.1.0", "type": "module", "scripts": { "dev": "node scripts/dev.js", "build": "vite build", "start": "node bin/setup.js", "init:env": "node bin/init-env.js", "doctor": "node bin/cli.js doctor" }, "dependencies": { "@frumu/tandem": "^0.4.7", "@frumu/tandem-client": "^0.4.7", "@tanstack/react-query": "^5.90.21", "dompurify": "^3.3.1", "lucide": "^0.575.0", "marked": "^17.0.3", "motion": "^11.18.2", "preact": "^10.28.4", "preact-router": "^4.1.2", "yaml": "^2.8.2" }, "devDependencies": { "@preact/preset-vite": "^2.10.5", "@tailwindcss/forms": "^0.5.11", "@types/react": "^19.2.14", "@types/react-dom": "^19.2.3", "autoprefixer": "^10.4.20"… 证据：`packages/create-tandem-panel/template/package.json` - **Package**（package_manifest）：{ "name": "tandem-ai", "version": "0.5.11", "description": "Tandem authority-layer runtime CLI installer/launcher", "license": "MIT", "bin": { "tandem": "bin/tandem.js" }, "publishConfig": { "access": "public", "git-checks": false }, "files": "bin", "scripts" } 证据：`packages/tandem-ai/package.json` - **Package**（package_manifest）：{ "name": "@frumu/tandem-client", "version": "0.5.11", "description": "TypeScript client for Tandem authority-layer runtime HTTP + SSE APIs", "type": "module", "main": "./dist/index.cjs", "module": "./dist/index.js", "types": "./dist/index.d.ts", "exports": { ".": { "types": "./dist/index.d.ts", "import": "./dist/index.js", "require": "./dist/index.cjs" } }, "scripts": { "build": "tsup src/index.ts --format esm,cjs --dts --clean", "dev": "tsup src/index.ts --format esm,cjs --dts --watch", "typecheck": "tsc --noEmit", "test": "vitest run", "prepublishOnly": "npm run build" }, "files": "dist" , "keywords": "tandem", "ai", "agent", "llm", "sse", "autonomous" , "license": "MIT", "repository": {… 证据：`packages/tandem-client-ts/package.json` - **Package**（package_manifest）：{ "name": "@frumu/tandem-panel", "version": "0.5.11", "description": "Web control center for Tandem authority-layer runtime: agents, workflows, memory, approvals, channels, and ops", "type": "module", "bin": { "tandem-control-panel": "bin/cli.js", "tandem-setup": "bin/cli.js", "tandem-control-panel-init": "bin/init-env.js" }, "scripts": { "build": "vite build", "dev": "vite", "preview": "vite preview", "test:smoke": "node --test tests/smoke.test.mjs", "init:env": "node bin/init-env.js", "doctor": "node bin/cli.js doctor", "docker:up": "docker compose up --build", "docker:down": "docker compose down", "docker:logs": "docker compose logs -f --tail=200", "docker:ps": "docker compose ps", "dock… 证据：`packages/tandem-control-panel/package.json` - **Package**（package_manifest）：{ "name": "@frumu/tandem", "version": "0.5.11", "description": "Tandem authority-layer runtime CLI and engine binary distribution", "homepage": "https://tandem.ac", "bin": { "tandem": "bin/tandem.js", "tandem-engine": "bin/tandem-engine.js" }, "scripts": { "postinstall": "node scripts/install.js" }, "files": "bin", "scripts" , "repository": { "type": "git", "url": "git+https://github.com/frumu-ai/tandem.git" }, "author": "Frumu Ltd", "license": "MIT OR Apache-2.0", "publishConfig": { "access": "public" }, "os": "darwin", "linux", "win32" , "cpu": "x64", "arm64" } 证据：`packages/tandem-engine/package.json` - **Package**（package_manifest）：{ "name": "@frumu/tandem-enterprise", "version": "0.5.11", "description": "Hosted enterprise Tandem engine binary distribution for Linux x64", "homepage": "https://tandem.ac", "bin": { "tandem-engine": "bin/tandem-engine.js" }, "scripts": { "postinstall": "node scripts/install.js" }, "files": "bin", "scripts" , "repository": { "type": "git", "url": "git+https://github.com/frumu-ai/tandem.git" }, "author": "Frumu Ltd", "license": "MIT OR Apache-2.0", "publishConfig": { "access": "public" }, "os": "linux" , "cpu": "x64" } 证据：`packages/tandem-enterprise/package.json` - **Package**（package_manifest）：{ "name": "@frumu/tandem-tui", "version": "0.5.11", "description": "Tandem authority-layer runtime TUI binary distribution", "homepage": "https://tandem.ac", "bin": { "tandem-tui": "bin/tandem-tui.js" }, "scripts": { "postinstall": "node scripts/install.js" }, "files": "bin", "scripts" , "repository": { "type": "git", "url": "git+https://github.com/frumu-ai/tandem.git" }, "author": "Frumu Ltd", "license": "MIT OR Apache-2.0", "publishConfig": { "access": "public" }, "os": "darwin", "linux", "win32" , "cpu": "x64", "arm64" } 证据：`packages/tandem-tui/package.json` - **Package**（package_manifest）：{ "name": "bench", "version": "1.0.0", "description": "", "main": "index.js", "scripts": { "test": "echo \"Error: no test specified\" && exit 1", "bench:server": "node bench server.mjs ../urls.txt", "bench:features": "node bench server features.mjs" }, "keywords": , "author": "", "license": "ISC", "dependencies": { "fast-glob": "^3.3.3", "jsdom": "^28.0.0", "p-limit": "^7.3.0", "turndown": "^7.2.2" } } 证据：`scripts/bench-js/package.json` - **AGENTS.md**（documentation）：Crate Purpose ----------------------- ------------------ crates/tandem-server/ Server application crates/tandem-core/ Core engine logic crates/tandem-cli/ CLI tools 证据：`AGENTS.md` - **Tandem Engine CLI Guide**（documentation）：This guide documents the master tandem CLI and the direct tandem-engine runtime using bash commands macOS/Linux/WSL . 证据：`docs/ENGINE_CLI.md` - **Engine Communication Guide**（documentation）：This document explains how Tandem clients communicate with tandem-engine , how runs stream, and how desktop/TUI coordinate engine lifecycle. 证据：`docs/ENGINE_COMMUNICATION.md` - **Engine Build, Run, and Test Guide**（documentation）：- how to build and start tandem-engine - how to run automated tests - how to run the end-to-end smoke/runtime proof flow on Windows, macOS, and Linux 证据：`docs/ENGINE_TESTING.md` - **Tandem Enterprise Readiness**（documentation）：This document separates what Tandem can credibly show today from what is still in progress or planned. Tandem is not yet a complete enterprise platform with full RBAC, OIDC, SCIM, SIEM export, SOC2, and private sidecar enforcement. The current proof is the runtime foundation those enterprise features will attach to. 证据：`docs/ENTERPRISE_READINESS.md` - **Tandem EU AI Act Readiness**（documentation）：Status: public readiness brief Date: 2026-05-19 Audience: CISOs, security reviewers, compliance teams, platform engineering 证据：`docs/EU_AI_ACT_COMPLIANCE.md` - **Tandem Licensing**（documentation）：Repository license class: Open Core / Mixed License Open-source components: MIT, Apache-2.0, or MIT OR Apache-2.0 Source-available components: Business Source License 1.1 BUSL-1.1 Default rule: Each package is governed by its package-local manifest and/or package-local license file. Canonical license map: This file. 证据：`docs/LICENSING.md` - **Role**（documentation）：You are a Staff Engineer / Product Architect agent. Your job is to design and plan and optionally implement a robust MCP “connector tools” integration for Tandem so that when a user connects an MCP server e.g., Arcade , Tandem automatically pulls the available tools, caches them, and lets users select/deselect which connector tools an agent is allowed to use in the Agent Builder UI. This must work for Arcade, Composio, and any other MCP-compatible connector service—no vendor-specific logic. 证据：`docs/MCP_IMPROVEMENTS.md` - **Release Notes**（documentation）：Canonical release notes now live at ../RELEASE NOTES.md ../RELEASE NOTES.md . 证据：`docs/RELEASE_NOTES.md` - **System Overview**（documentation）：This page gives a practical mental model of how Tandem components fit together. 证据：`guide/src/content/docs/architecture.md` - **What this flow proves**（documentation）：Use this guide when you want the fastest path from human intent to a real automationsV2 payload. 证据：`guide/src/content/docs/automation-composer-workflows.md` - **🖥️ I am a Desktop User**（documentation）：Welcome to the Tandem documentation hub . Tandem is the authority layer for AI-first work: an engine-owned runtime that controls what agents can see, which tools they can use, when humans must approve, and what evidence survives after the work is done. 证据：`guide/src/content/docs/index.md` - **What You Get**（documentation）：Set up scheduled agents that can use MCP connector tools with explicit per-agent tool allowlists. 证据：`guide/src/content/docs/mcp-automated-agents.md` - **The Four States**（documentation）：Use this guide when an agent needs to answer a capability question safely: 证据：`guide/src/content/docs/mcp-capability-discovery-and-request-flow.md` - **Prerequisites**（documentation）：import { LinkCard, CardGrid } from "@astrojs/starlight/components"; 证据：`guide/src/content/docs/sdk/index.md` - **Install**（documentation）：pip install tandem-client installs the Python SDK only. It does not install tandem-engine . 证据：`guide/src/content/docs/sdk/python.md` - **Install**（documentation）：Requires Node.js 18+ uses native fetch and ReadableStream . 证据：`guide/src/content/docs/sdk/typescript.md` - **Contributing to Tandem**（documentation）：Thank you for your interest in contributing to Tandem! This document provides guidelines and instructions for contributing. 证据：`CONTRIBUTING.md` - **Contributing Bio-Informatics Pack**（documentation）：Do not use global pip install ... instructions in this pack's docs. 证据：`agent-templates/pack-docs/bio-informatics-pack/CONTRIBUTING.md` - **Contributing Data Visualization Pack**（documentation）：Contributing Data Visualization Pack 证据：`agent-templates/pack-docs/data-visualization-pack/CONTRIBUTING.md` - **Contributing Finance Analysis Pack**（documentation）：This pack includes Python templates. Do not recommend global pip install ... in docs. 证据：`agent-templates/pack-docs/finance-analysis-pack/CONTRIBUTING.md` - **Bio-Informatics Skill**（skill_instruction）：Run a reproducible bioinformatics workflow from source preparation to analysis-ready deliverables. 证据：`agent-templates/packs/bio-informatics-pack/skills/bio-informatics-analyst/SKILL.md` - **Data Visualization Skill**（skill_instruction）：Create accurate, publication-ready visuals that communicate a clear analytical story. 证据：`agent-templates/packs/data-visualization-pack/skills/data-visualization-analyst/SKILL.md` - **Finance Analysis Skill**（skill_instruction）：Turn financial inputs into reliable variance analysis and executive-ready recommendations. 证据：`agent-templates/packs/finance-analysis-pack/skills/finance-analysis-analyst/SKILL.md` - **Legal Research Skill**（skill_instruction）：Analyze legal materials with consistent issue spotting, risk ranking, and concise legal writing. 证据：`agent-templates/packs/legal-research-pack/skills/legal-research-analyst/SKILL.md` - **Micro-Drama Script Studio Skill**（skill_instruction）：Turn the pack inputs into a publishable short episode package with strong continuity and clear production handoff. 证据：`agent-templates/packs/micro-drama-script-studio-pack/skills/micro-drama-script-studio/SKILL.md` - **Research Synthesis Skill**（skill_instruction）：Produce a clear, defensible synthesis from many sources while preserving uncertainty and conflicts. 证据：`agent-templates/packs/research-synthesis-pack/skills/research-synthesis-analyst/SKILL.md` - 其余 118 条证据见 `AI_CONTEXT_PACK.json` 或 `EVIDENCE_INDEX.json`。 ## 宿主 AI 必须遵守的规则 - **把本资产当作开工前上下文，而不是运行环境。**：AI Context Pack 只包含证据化项目理解，不包含目标项目的可执行状态。 证据：`docs/README.md`, `docs/compliance/README.md`, `.agents/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：`docs/README.md`, `docs/compliance/README.md`, `.agents/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。 - **Tandem 概述**：importance `high` - source_paths: README.md, VISION.md, PRODUCT.md, PRINCIPLES.md - **系统架构**：importance `high` - source_paths: ARCHITECTURE.md, contracts/events.json, contracts/http.md, manifests/components/tandem-engine.yaml, manifests/components/tandem-desktop.yaml - **核心概念**：importance `high` - source_paths: crates/tandem-core/src/engine_loop/mod.rs, crates/tandem-core/src/permissions.rs, crates/tandem-core/src/tool_policy.rs, crates/tandem-memory/src/lib.rs, crates/tandem-memory/src/types.rs - **快速开始**：importance `high` - source_paths: README.md, docs/ENGINE_CLI.md, packages/tandem-control-panel/README.md, packages/tandem-control-panel/DOCKER.md, guide/src/content/docs/installation.md - **桌面应用**：importance `high` - source_paths: apps/tandem-desktop/src-tauri/src/lib.rs, apps/tandem-desktop/src-tauri/tauri.conf.json, guide/src/content/docs/desktop/overview.md, guide/src/content/docs/desktop/first-10-minutes.md, guide/src/content/docs/desktop/workflows.md - **引擎 CLI 与无头服务**：importance `high` - source_paths: docs/ENGINE_CLI.md, docs/ENGINE_COMMUNICATION.md, docs/ENGINE_TESTING.md, engine/src/main.rs, crates/tandem-server/src/http/router.rs - **SDK 参考文档**：importance `high` - source_paths: packages/tandem-client-ts/src/client.ts, packages/tandem-client-ts/src/stream.ts, packages/tandem-client-py/tandem_client/client.py, packages/tandem-client-py/tandem_client/stream.py, packages/tandem-client-ts/README.md - **MCP 集成**：importance `high` - source_paths: crates/tandem-runtime/src/mcp.rs, crates/tandem-runtime/src/mcp_ready.rs, crates/tandem-server/src/mcp_catalog.rs, crates/tandem-server/resources/mcp-catalog/index.json, docs/MCP_IMPROVEMENTS.md ## Repo Inspection Evidence / 源码检查证据 - repo_clone_verified: true - repo_inspection_verified: true - repo_commit: `2b32b69382fe9e9b968add778187f5537e657d90` - inspected_files: `package.json`, `README.md`, `docs/WORKFLOW_BUG_REPLAY.md`, `docs/DESIGN_SYSTEM.md`, `docs/README_TRANSLATIONS.md`, `docs/ENGINE_CLI.md`, `docs/marketing_skill_canonical_map.md`, `docs/SMART_HEARTBEAT_PATTERN.md`, `docs/ENGINE_PROTOCOL_MATRIX.md`, `docs/LICENSING.md`, `docs/OLLAMA_GUIDE.md`, `docs/PLATFORM_COMPATIBILITY.md`, `docs/ENGINE_TESTING.md`, `docs/I18N_GUIDE.md`, `docs/EU_AI_ACT_COMPLIANCE.md`, `docs/WORKFLOW_GENERATED_VARIATIONS.md`, `docs/LOGGING_SCHEMA.md`, `docs/bug-monitor-external-log-intake-demo.md`, `docs/ENTERPRISE_READINESS.md`, `docs/AI_RUNTIME_INFRASTRUCTURE.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: 来源证据：Workflow automation-v2-e42950f5-9d9e-483c-a8da-3d568f98dd46 failed at research_sources: automation run blocked by upstr… - Trigger: GitHub 社区证据显示该项目存在一个安装相关的待验证问题：Workflow automation-v2-e42950f5-9d9e-483c-a8da-3d568f98dd46 failed at research_sources: automation run blocked by upstream node outcome - 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: 可能阻塞安装或首次运行。 - Evidence: community_evidence:github | cevd_d7998fa92e2f4cae91e7040b71c38bf9 | https://github.com/frumu-ai/tandem/issues/38 | 来源讨论提到 node 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 2: 来源证据：Workflow automation-v2-1f391975-681a-4041-94b5-8845dfe3307b failed at execute_goal: automation run blocked by upstream… - Trigger: GitHub 社区证据显示该项目存在一个配置相关的待验证问题：Workflow automation-v2-1f391975-681a-4041-94b5-8845dfe3307b failed at execute_goal: automation run blocked by upstream node outcome - 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: 可能阻塞安装或首次运行。 - Evidence: community_evidence:github | cevd_b3856a6aadb944f0977020d9841025a5 | https://github.com/frumu-ai/tandem/issues/28 | 来源讨论提到 node 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 3: 来源证据：Workflow automation-v2-65662b51-ab4e-4348-8349-ea10e7b634c6 failed at synthesize_report: automation run failed from nod… - Trigger: GitHub 社区证据显示该项目存在一个配置相关的待验证问题：Workflow automation-v2-65662b51-ab4e-4348-8349-ea10e7b634c6 failed at synthesize_report: automation run failed from node outcomes: synthesize_report - 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_8c10aeb4b1874bf39ce9347480864a20 | https://github.com/frumu-ai/tandem/issues/1396 | 来源讨论提到 node 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 4: 来源证据：Workflow automation-v2-7a7374b2-16bc-416d-bc38-81edb5be7050 failed at assess: automation run failed from node outcomes:… - Trigger: GitHub 社区证据显示该项目存在一个配置相关的待验证问题：Workflow automation-v2-7a7374b2-16bc-416d-bc38-81edb5be7050 failed at assess: automation run failed from node outcomes: assess - 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_8fda7ba929c344aa863117d5164cb1e5 | https://github.com/frumu-ai/tandem/issues/32 | 来源讨论提到 node 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 5: 来源证据：Workflow automation-v2-9ee33834-bf6d-4f86-acb3-3cd41d9cef19 failed at assess_reddit_activity: automation run blocked by… - Trigger: GitHub 社区证据显示该项目存在一个配置相关的待验证问题：Workflow automation-v2-9ee33834-bf6d-4f86-acb3-3cd41d9cef19 failed at assess_reddit_activity: automation run blocked by upstream node outcome - 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: 可能阻塞安装或首次运行。 - Evidence: community_evidence:github | cevd_55cc9e60475944d78cd5af4305d0f1ef | https://github.com/frumu-ai/tandem/issues/35 | 来源讨论提到 node 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 6: 来源证据：Workflow automation-v2-9ee33834-bf6d-4f86-acb3-3cd41d9cef19 failed at search_agents_keep: automation node `search_agent… - Trigger: GitHub 社区证据显示该项目存在一个配置相关的待验证问题：Workflow automation-v2-9ee33834-bf6d-4f86-acb3-3cd41d9cef19 failed at search_agents_keep: automation node `search_agents_keep` timed out after 180000 ms - 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_fd1ae3d40b5d4c97b8e86767f85466b4 | https://github.com/frumu-ai/tandem/issues/40 | 来源讨论提到 node 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 7: 来源证据：Workflow automation-v2-9ee33834-bf6d-4f86-acb3-3cd41d9cef19 failed at search_multi_agent: automation node `search_multi… - Trigger: GitHub 社区证据显示该项目存在一个配置相关的待验证问题：Workflow automation-v2-9ee33834-bf6d-4f86-acb3-3cd41d9cef19 failed at search_multi_agent: automation node `search_multi_agent` timed out after 180000 ms - 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_9f60d02de0544ffcbacc2c984db5ccaf | https://github.com/frumu-ai/tandem/issues/36 | 来源讨论提到 node 相关条件，需在安装/试用前复核。 - Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。 ### Constraint 8: 来源证据：Workflow automation-v2-a776164f-06ef-440d-af54-748186205aef failed at analyze_findings: provider stream chunk error: er… - Trigger: GitHub 社区证据显示该项目存在一个配置相关的待验证问题：Workflow automation-v2-a776164f-06ef-440d-af54-748186205aef failed at analyze_findings: provider stream chunk error: error decoding response body - 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_7f64d78b216242bd8b6108520a197aff | https://github.com/frumu-ai/tandem/issues/49 | 来源讨论提到 node 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 9: 来源证据：Workflow automation-v2-a776164f-06ef-440d-af54-748186205aef failed at research_sources: automation run blocked by upstr… - Trigger: GitHub 社区证据显示该项目存在一个配置相关的待验证问题：Workflow automation-v2-a776164f-06ef-440d-af54-748186205aef failed at research_sources: automation run blocked by upstream node outcome - 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: 可能阻塞安装或首次运行。 - Evidence: community_evidence:github | cevd_bb5c682d661c42d7b568ae17b9e8e0f9 | https://github.com/frumu-ai/tandem/issues/41 | 来源讨论提到 node 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 10: 来源证据：Workflow automation-v2-b2d6ef35-c222-4027-9251-40adca379cf3 failed at collect_recent_files: required output `.tandem/ru… - Trigger: GitHub 社区证据显示该项目存在一个配置相关的待验证问题：Workflow automation-v2-b2d6ef35-c222-4027-9251-40adca379cf3 failed at collect_recent_files: required output `.tandem/runs/automation-v2-run-593051dc-78bf-4927-… - 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_fed96ae76b684a42b4e5a90f3f421837 | https://github.com/frumu-ai/tandem/issues/34 | 来源讨论提到 node 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。