# openai-agents-js - 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 openai-agents-js 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 - **希望把专业流程带进宿主 AI 的用户**：仓库包含 Skill 文档。 证据：`.agents/skills/changeset-validation/SKILL.md`, `.agents/skills/code-change-verification/SKILL.md`, `.agents/skills/docs-sync/SKILL.md`, `.agents/skills/examples-auto-run/SKILL.md` 等 Claim：`clm_0003` 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：`.agents/skills/changeset-validation/SKILL.md`, `.agents/skills/code-change-verification/SKILL.md`, `.agents/skills/docs-sync/SKILL.md`, `.agents/skills/examples-auto-run/SKILL.md` 等 Claim：`clm_0001` supported 0.86 - **命令行启动或安装流程**（需要安装后验证）：项目文档中存在可执行命令，真实使用需要在本地或宿主环境中运行这些命令。 证据：`README.md`, `packages/agents-core/README.md`, `packages/agents-extensions/README.md` Claim：`clm_0002` supported 0.86 ## 怎么开始 - `npm install @openai/agents zod` 证据：`README.md` Claim：`clm_0004` supported 0.86 - `npm install @openai/agents` 证据：`packages/agents-core/README.md` Claim：`clm_0004` supported 0.86, `clm_0005` supported 0.86, `clm_0006` supported 0.86 - `npm install @openai/agents @openai/agents-extensions` 证据：`packages/agents-extensions/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：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：`.agents/skills/changeset-validation/SKILL.md`, `.agents/skills/code-change-verification/SKILL.md`, `.agents/skills/docs-sync/SKILL.md`, `.agents/skills/examples-auto-run/SKILL.md` 等 Claim：`clm_0003` 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：`.agents/skills/changeset-validation/SKILL.md`, `.agents/skills/code-change-verification/SKILL.md`, `.agents/skills/docs-sync/SKILL.md`, `.agents/skills/examples-auto-run/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`, `packages/agents-core/README.md`, `packages/agents-extensions/README.md` Claim：`clm_0002` supported 0.86 - **存在 Quick Start / 安装命令线索**（supported）：可以相信项目文档出现过启动或安装入口；不要因此直接在主力环境运行。 证据：`README.md` Claim：`clm_0004` 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：`.agents/skills/changeset-validation/SKILL.md`, `.agents/skills/code-change-verification/SKILL.md`, `.agents/skills/docs-sync/SKILL.md`, `.agents/skills/examples-auto-run/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。 - **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`, `packages/agents-core/README.md`, `packages/agents-extensions/README.md` - **宿主 AI 配置**：Claude/Codex/Cursor/Gemini/OpenCode 等宿主的 plugin、Skill 或规则加载配置。 原因：宿主配置会改变 AI 后续工作方式，可能和用户已有规则冲突。 证据：`.agents/skills/changeset-validation/SKILL.md`, `.agents/skills/code-change-verification/SKILL.md`, `.agents/skills/docs-sync/SKILL.md`, `.agents/skills/examples-auto-run/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：`README.md`, `packages/agents-core/README.md`, `packages/agents-extensions/README.md` - **宿主 AI 上下文**：AI Context Pack、Prompt Preview、Skill 路由、风险规则和项目事实。 原因：导入上下文会影响宿主 AI 后续判断，必须避免把未验证项包装成事实。 ### 最小安全下一步 - **先跑 Prompt Preview**：用安装前交互式试用判断工作方式是否匹配，不需要授权或改环境。（适用：任何项目都适用，尤其是输出质量未知时。） - **只在隔离目录或测试账号试装**：避免安装命令污染主力宿主 AI、真实项目或用户主目录。（适用：存在命令执行、插件配置或本地写入线索时。） - **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 - **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`, `packages/agents-core/README.md`, `packages/agents-extensions/README.md` Claim：`clm_0008` supported 0.86 - **待确认**：真实安装后是否与用户当前宿主 AI 版本兼容？。原因：兼容性只能通过实际宿主环境验证。 - **待确认**：项目输出质量是否满足用户具体任务？。原因：安装前预览只能展示流程和边界，不能替代真实评测。 - **待确认**：安装命令是否需要网络、权限或全局写入？。原因：这影响企业环境和个人环境的安装风险。 ## 开工前工作上下文 ### 加载顺序 - 先读取 how_to_use.host_ai_instruction，建立安装前判断资产的边界。 - 读取 claim_graph_summary，确认事实来自 Claim/Evidence Graph，而不是 Human Wiki 叙事。 - 再读取 intended_users、capabilities 和 quick_start_candidates，判断用户是否匹配。 - 需要执行具体任务时，优先查 role_skill_index，再查 evidence_index。 - 遇到真实安装、文件修改、网络访问、性能或兼容性问题时，转入 risk_card 和 boundaries.runtime_required。 ### 任务路由 - **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：`.agents/skills/changeset-validation/SKILL.md`, `.agents/skills/code-change-verification/SKILL.md`, `.agents/skills/docs-sync/SKILL.md`, `.agents/skills/examples-auto-run/SKILL.md` 等 Claim：`clm_0001` supported 0.86 - **命令行启动或安装流程**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`README.md`, `packages/agents-core/README.md`, `packages/agents-extensions/README.md` Claim：`clm_0002` supported 0.86 ### 上下文规模 - 文件总数：1210 - 重要文件覆盖：40/1210 - 证据索引条目：86 - 角色 / Skill 条目：15 ### 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 openai-agents-js 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 openai-agents-js 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 openai-agents-js 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 15 个角色 / Skill / 项目文档条目。 - **changeset-validation**（skill）：Validate changesets in openai-agents-js using LLM judgment against git diffs including uncommitted local changes . Use when packages/ or .changeset/ are modified, or when verifying PR changeset compliance and bump level. 激活提示：当用户任务与“changeset-validation”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：`.agents/skills/changeset-validation/SKILL.md` - **code-change-verification**（skill）：Run the mandatory verification stack when changes affect runtime code, tests, or build/test behavior in the OpenAI Agents JS monorepo. 激活提示：当用户任务与“code-change-verification”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：`.agents/skills/code-change-verification/SKILL.md` - **docs-sync**（skill）：Analyze main branch implementation and configuration to find missing, incorrect, or outdated documentation in docs/. Use when asked to audit doc coverage, sync docs with code, or propose doc updates/structure changes. Only update English docs docs/src/content/docs/ and never touch translated docs under docs/src/content/docs/ja, ko, or zh. Provide a report and ask for approval before editing docs. 激活提示：当用户任务与“docs-sync”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：`.agents/skills/docs-sync/SKILL.md` - **examples-auto-run**（skill）：Run examples:start-all in auto mode with parallel execution, per-script logs, and start/stop helpers. 激活提示：当用户任务与“examples-auto-run”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：`.agents/skills/examples-auto-run/SKILL.md` - **final-release-review**（skill）：Perform a release-readiness review by locating the previous release tag from remote tags and auditing the diff e.g., v1.2.3... for breaking changes, regressions, improvement opportunities, and risks before releasing openai-agents-js. 激活提示：当用户任务与“final-release-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：`.agents/skills/final-release-review/SKILL.md` - **implementation-strategy**（skill）：Decide how to implement runtime and API changes in openai-agents-js before editing code. Use when a task changes exported APIs, runtime behavior, schemas, tests, or docs and you need to choose the compatibility boundary, whether shims or migrations are warranted, and when unreleased interfaces can be rewritten directly. 激活提示：当用户任务与“implementation-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：`.agents/skills/implementation-strategy/SKILL.md` - **integration-tests**（skill）：Run the integration-tests pipeline that depends on a local npm registry Verdaccio . Use when asked to execute integration tests or local publish workflows in this repo. 激活提示：当用户任务与“integration-tests”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：`.agents/skills/integration-tests/SKILL.md` - **openai-knowledge**（skill）：Use when working with the OpenAI API Responses API or OpenAI platform features tools, streaming, Realtime API, auth, models, rate limits, MCP and you need authoritative, up-to-date documentation schemas, examples, limits, edge cases . Prefer the OpenAI Developer Documentation MCP server tools when available; otherwise guide the user to enable openaiDeveloperDocs . 激活提示：当用户任务与“openai-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：`.agents/skills/openai-knowledge/SKILL.md` - **pnpm-upgrade**（skill）：Keep pnpm current: run pnpm self-update/corepack prepare, align packageManager in package.json, and bump pnpm/action-setup + pinned pnpm versions in .github/workflows to the latest release. Use this when refreshing the pnpm toolchain manually or in automation. 激活提示：当用户任务与“pnpm-upgrade”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：`.agents/skills/pnpm-upgrade/SKILL.md` - **pr-draft-summary**（skill）：Create the required PR-ready summary block, branch suggestion, title, and draft description for openai-agents-js. Use in the final handoff after moderate-or-larger changes to runtime code, tests, examples, build/test configuration, or docs with behavior impact; skip only for trivial or conversation-only tasks, repo-meta/doc-only tasks without behavior impact, or when the user explicitly says not to include the PR dr… 激活提示：当用户任务与“pr-draft-summary”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：`.agents/skills/pr-draft-summary/SKILL.md` - **runtime-behavior-probe**（skill）：Plan and execute runtime-behavior investigations with temporary TypeScript probe scripts, validation matrices, state controls, and findings-first reports. Use only when the user explicitly invokes this skill to verify actual runtime behavior beyond normal code-level checks, especially to uncover edge cases, undocumented behavior, or common failure modes in local or live integrations. A baseline smoke check is fine a… 激活提示：当用户任务与“runtime-behavior-probe”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：`.agents/skills/runtime-behavior-probe/SKILL.md` - **test-coverage-improver**（skill）：Improve test coverage in the OpenAI Agents JS monorepo: run pnpm test:coverage , inspect coverage artifacts, identify low-coverage files and branches, propose high-impact tests, and confirm with the user before writing tests. 激活提示：当用户任务与“test-coverage-improver”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：`.agents/skills/test-coverage-improver/SKILL.md` - **invoice-total-fixer**（skill）：Use when fixing invoice total calculations in the sandbox quickstart repository. 激活提示：当用户任务与“invoice-total-fixer”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：`examples/docs/sandbox-agents/skills/invoice-total-fixer/SKILL.md` - **credit-note-fixer**（skill）：Fix the tiny credit-note formatting bug and rerun the exact targeted test command. 激活提示：当用户任务与“credit-note-fixer”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：`examples/sandbox/data/coding-task/skills/credit-note-fixer/SKILL.md` - **csv-workbench**（skill）：Analyze CSV files in /mnt/data and return concise numeric summaries. 激活提示：当用户任务与“csv-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：`examples/tools/skills/csv-workbench/SKILL.md` ## 证据索引 - 共索引 86 条证据。 - **Docs**（documentation）：The documentation is generated using Astro Starlight. 证据：`docs/README.md` - **Documentation Snippets**（documentation）：This directory contains small scripts used throughout the documentation. Run them with pnpm using the commands shown below. 证据：`examples/docs/README.md` - **Invoice Total Fixer**（skill_instruction）：Inspect the implementation and test before editing. The total should apply tax as a percentage of the subtotal, so the expected formula is subtotal + subtotal taxRate . 证据：`examples/docs/sandbox-agents/skills/invoice-total-fixer/SKILL.md` - **Changesets**（documentation）：Hello and welcome! This folder has been automatically generated by @changesets/cli , a build tool that works with multi-package repos, or single-package repos to help you version and publish your code. You can find the full documentation for it in our repository https://github.com/changesets/changesets 证据：`.changeset/README.md` - **Contributor Guide**（documentation）：This guide helps new contributors get started with the OpenAI Agents JS monorepo. It covers repo structure, how to test your work, available utilities, file locations, and guidelines for commits and PRs. 证据：`AGENTS.md` - **OpenAI Agents SDK JavaScript/TypeScript**（documentation）：OpenAI Agents SDK JavaScript/TypeScript 证据：`README.md` - **Integration tests**（documentation）：This project hosts packages to test the different environments that the Agents SDK works in. 证据：`integration-tests/README.md` - **Agent Pattern Examples**（documentation）：This directory contains small scripts that demonstrate different agent patterns. Run them with pnpm using the commands shown below. 证据：`examples/agent-patterns/README.md` - **AI SDK UI Stream Example**（documentation）：This example shows how to convert an Agents SDK streaming run into responses that are compatible with the AI SDK UI data stream and text stream protocols. 证据：`examples/ai-sdk-ui/README.md` - **AI SDK Example**（documentation）：This example shows how to run the Agents SDK with a model provided by the AI SDK https://www.npmjs.com/package/@ai-sdk/openai . 证据：`examples/ai-sdk-v1/README.md` - **AI SDK Examples**（documentation）：These examples show how to wrap models from the AI SDK https://www.npmjs.com/package/@ai-sdk/openai and compatible providers with the aisdk helper from @openai/agents-extensions/ai-sdk , then run them inside the Agents runtime. 证据：`examples/ai-sdk/README.md` - **Basic Examples**（documentation）：This directory contains small scripts that demonstrate features of the Agents SDK. Run them with pnpm using the commands shown below. 证据：`examples/basic/README.md` - **Customer Service Agent**（documentation）：This example demonstrates a multi-agent customer service workflow for an airline. The index.ts script sets up a triage agent that can delegate to specialized FAQ and seat booking agents. Tools are used to look up common questions and to update a passenger's seat. Interaction occurs through a simple CLI loop, showing how agents can hand off between each other and call tools. 证据：`examples/customer-service/README.md` - **Financial Research Agent**（documentation）：This example demonstrates a multi-agent workflow that produces a short financial analysis report. 证据：`examples/financial-research-agent/README.md` - **Agent Handoffs**（documentation）：This example shows how one agent can transfer control to another. The index.ts script sets up two English speaking assistants and a Spanish assistant. The second agent is configured with a handoff so that if the user requests Spanish replies it hands off to the Spanish agent. A message filter strips out tool messages and the first two history items before the handoff occurs. Run it with: 证据：`examples/handoffs/README.md` - **Model Context Protocol Example**（documentation）：This example demonstrates how to use the Model Context Protocol https://modelcontextprotocol.io/ with the OpenAI Agents SDK. 证据：`examples/mcp/README.md` - **Model Providers Examples**（documentation）：This directory contains small scripts showing how to integrate custom model providers. Run them with pnpm using the commands shown below. 证据：`examples/model-providers/README.md` - **Next.js Demo**（documentation）：This example shows a basic example of how to use human-in-the-loop in a Next.js application. 证据：`examples/nextjs/README.md` - **Realtime Demo**（documentation）：This example is a small Vite https://vitejs.dev/ application showcasing the realtime agent API. 证据：`examples/realtime-demo/README.md` - **Realtime Next.js Demo**（documentation）：This example shows how to combine Next.js with the OpenAI Agents SDK to create a realtime voice agent. 证据：`examples/realtime-next/README.md` - **Twilio SIP Realtime Example**（documentation）：This example shows how to handle OpenAI Realtime SIP calls with the Agents JS SDK. Incoming calls are accepted through the Realtime Calls API, a triage agent answers with a fixed greeting, and handoffs route the caller to specialist agents FAQ lookup and record updates similar to the realtime UI demo. 证据：`examples/realtime-twilio-sip/README.md` - **Realtime Twilio Integration**（documentation）：This example demonstrates how to connect the OpenAI Realtime API to a phone call using Twilio's Media Streams. The script in index.ts starts a Fastify server that serves TwiML for incoming calls and creates a WebSocket endpoint for streaming audio. When a call connects, the audio stream is forwarded through a TwilioRealtimeTransportLayer to a RealtimeSession so the RealtimeAgent can respond in real time. 证据：`examples/realtime-twilio/README.md` - **Research Bot**（documentation）：This example shows how to orchestrate several agents to produce a detailed research report. 证据：`examples/research-bot/README.md` - **Sandbox Examples**（documentation）：These examples show the JavaScript sandbox APIs that are implemented in this branch: Manifest , SandboxAgent , local and Docker sandbox clients, filesystem and shell capabilities, lazy skills, host tools, handoffs, local and remote snapshots, external memory stores, and SDK conversation sessions. 证据：`examples/sandbox/README.md` - **Credit Note Example Repo**（documentation）：This tiny repo exists to support examples/sandbox/coding-task.ts . 证据：`examples/sandbox/data/coding-task/repo/README.md` - **Cloud Sandbox Extension Examples**（documentation）：These examples mirror the Python sandbox extension runners and keep the same environment-variable names and flag shapes where that makes sense. 证据：`examples/sandbox/extensions/README.md` - **Tool Integrations**（documentation）：These examples demonstrate the hosted tools provided by the Agents SDK. 证据：`examples/tools/README.md` - **OpenAI Agents SDK**（documentation）：The OpenAI Agents SDK is a lightweight yet powerful framework for building multi-agent workflows. 证据：`packages/agents-core/README.md` - **OpenAI Agents SDK Extensions**（documentation）：This package contains a collection of extension features for the OpenAI Agents SDK and is intended to be used alongside it. 证据：`packages/agents-extensions/README.md` - **OpenAI Agents SDK**（documentation）：The OpenAI Agents SDK is a lightweight yet powerful framework for building multi-agent workflows. 证据：`packages/agents-openai/README.md` - **OpenAI Agents SDK**（documentation）：The OpenAI Agents SDK is a lightweight yet powerful framework for building multi-agent workflows. 证据：`packages/agents-realtime/README.md` - **OpenAI Agents SDK JavaScript/TypeScript**（documentation）：OpenAI Agents SDK JavaScript/TypeScript 证据：`packages/agents/README.md` - **Package**（package_manifest）：{ "private": true, "name": "docs", "type": "module", "scripts": { "dev": "astro dev", "start": "astro dev", "build": "astro build", "preview": "astro preview", "astro": "astro", "translate": "tsx src/scripts/translate.ts" }, "dependencies": { "@astrojs/starlight": "^0.38.4", "@astrojs/starlight-tailwind": "^5.0.0", "@openai/agents": "workspace: ", "@tailwindcss/vite": "^4.2.4", "astro": "^6.1.10", "sharp": "^0.34.5", "starlight-llms-txt": "^0.8.1", "starlight-typedoc": "^0.21.5", "typedoc": "^0.28.19", "typedoc-plugin-markdown": "^4.11.0" }, "devDependencies": { "tailwindcss": "^4.2.4", "tsx": "^4.21.0", "typedoc-plugin-frontmatter": "^1.3.1", "typedoc-plugin-zod": "^1.4.3" } } 证据：`docs/package.json` - **Package**（package_manifest）：{ "private": true, "name": "openai-agents-js", "version": "0.0.1", "scripts": { "clean": "node scripts/clean-package-dists.mjs && tsc-multi --clean", "prebuild": "pnpm -F @openai/ -r prebuild", "build": "pnpm clean && tsc-multi", "build:ci": "pnpm run prebuild && pnpm clean && tsc-multi --maxWorkers 1 && pnpm run postbuild", "postbuild": "pnpm -r -F @openai/ bundle", "packages:dev": "tsc-multi --watch", "docs:dev": "pnpm -F docs dev", "docs:translate": "pnpm -F docs translate", "docs:build": "pnpm -F docs build", "docs:scripts:check": "pnpm exec tsc --pretty false --project docs/tsconfig.scripts.json", "test": "CI=1 NODE ENV=test vitest", "test:coverage": "NODE ENV=test vitest run --coverag… 证据：`package.json` - **Contributing to OpenAI Agents SDK**（documentation）：Thank you for your interest in contributing to the OpenAI Agents SDK. This document outlines the process for reporting issues, proposing changes, and submitting pull requests. 证据：`CONTRIBUTING.md` - **Package**（package_manifest）：{ "private": true, "name": "agent-patterns", "dependencies": { "@openai/agents": "workspace: ", "chalk": "^5.4.1", "zod": "^4.0.0" }, "scripts": { "build-check": "tsc --noEmit", "start:agents-as-tools": "tsx agents-as-tools.ts", "start:agents-as-tools-conditional": "tsx agents-as-tools-conditional.ts", "start:agents-as-tools-structured": "tsx agents-as-tools-structured.ts", "start:agents-as-tools-streaming": "tsx agents-as-tools-streaming.ts", "start:deterministic": "tsx deterministic.ts", "start:forcing-tool-use": "tsx forcing-tool-use.ts -t default", "start:human-in-the-loop-stream": "tsx human-in-the-loop-stream.ts", "start:human-in-the-loop": "tsx human-in-the-loop.ts", "start:input-gua… 证据：`examples/agent-patterns/package.json` - **Package**（package_manifest）：{ "private": true, "name": "ai-sdk-ui", "dependencies": { "@ai-sdk/react": "^3.0.44", "@openai/agents": "workspace: ", "@openai/agents-extensions": "workspace: ", "@openai/agents-openai": "workspace: ", "ai": "^6.0.42", "next": "16.2.3", "react": "^19.2.3", "react-dom": "^19.2.3", "react-markdown": "^9.0.3", "rehype-sanitize": "^6.0.0", "remark-gfm": "^4.0.0", "zod": "^4.0.0" }, "scripts": { "build-check": "tsc --noEmit", "dev": "next dev --turbopack", "build": "next build", "start": "next start", "lint": "next lint", "start:script-simple": "tsx scripts/simple.ts", "start:script-ai-sdk": "tsx scripts/ai-sdk.ts" }, "devDependencies": { "@ai-sdk/openai": "^3.0.14", "@types/node": "^20", "@typ… 证据：`examples/ai-sdk-ui/package.json` - **Package**（package_manifest）：{ "private": true, "name": "ai-sdk-v1", "dependencies": { "@ai-sdk/openai": "1", "@ai-sdk/provider": "1.1.3", "@openai/agents": "0.3.9", "@openai/agents-extensions": "0.3.9", "hono": "^4.12.16", "zod": "^3.23.8" }, "scripts": { "build-check": "tsc --noEmit", "start": "tsx index.ts", "start:stream": "tsx stream.ts" } } 证据：`examples/ai-sdk-v1/package.json` - **Package**（package_manifest）：{ "private": true, "name": "ai-sdk", "dependencies": { "@ai-sdk/anthropic": "^3.0.9", "@ai-sdk/google": "^3.0.6", "@ai-sdk/openai": "^3.0.7", "@openai/agents": "workspace: ", "@openai/agents-extensions": "workspace: ", "@openrouter/ai-sdk-provider": "^2.9.0", "zod": "4.3.5" }, "scripts": { "build-check": "tsc --noEmit", "start": "tsx index.ts", "start:gpt-5": "tsx gpt-5.ts", "start:retry": "tsx retry.ts", "start:stream": "tsx stream.ts", "start:image-tool-output": "tsx image-tool-output.ts" }, "devDependencies": { "ai": "^6.0.3" } } 证据：`examples/ai-sdk/package.json` - **Package**（package_manifest）：{ "private": true, "name": "basic", "dependencies": { "@openai/agents": "workspace: ", "openai": "^6.35.0", "zod": "^4.0.0" }, "scripts": { "build-check": "tsc --noEmit", "start": "tsx index.ts", "start:agent-lifecycle-example": "tsx agent-lifecycle-example.ts", "start:chat": "tsx chat.ts", "start:dynamic-system-prompt": "tsx dynamic-system-prompt.ts", "start:hello-world": "tsx hello-world.ts", "start:hello-world-gpt-5": "tsx hello-world-gpt-5.ts", "start:hello-world-gpt-oss": "tsx hello-world-gpt-oss.ts", "start:lifecycle-example": "tsx lifecycle-example.ts", "start:local-image": "tsx local-image.ts", "start:image-tool-output": "tsx image-tool-output.ts", "start:file-tool-output": "tsx fil… 证据：`examples/basic/package.json` - **Package**（package_manifest）：{ "private": true, "name": "connectors", "dependencies": { "@openai/agents": "workspace: " }, "scripts": { "build-check": "tsc --noEmit", "start": "tsx index.ts" } } 证据：`examples/connectors/package.json` - **Package**（package_manifest）：{ "private": true, "name": "customer-service", "dependencies": { "@openai/agents": "workspace: ", "@openai/agents-core": "workspace: ", "zod": "^4.0.0" }, "scripts": { "build-check": "tsc --noEmit", "start": "tsx index.ts" } } 证据：`examples/customer-service/package.json` - **Package**（package_manifest）：{ "private": true, "name": "docs-code", "dependencies": { "@ai-sdk/openai": "^2.0.15", "@modelcontextprotocol/server-filesystem": "^2026.1.14", "@openai/agents": "workspace: ", "@openai/agents-core": "workspace: ", "@openai/agents-extensions": "workspace: ", "@openai/agents-realtime": "workspace: ", "openai": "^6.35.0", "server-only": "^0.0.1", "zod": "^4.0.0" }, "scripts": { "build-check": "tsc --noEmit" }, "devDependencies": { "typedoc-plugin-zod": "^1.4.1" } } 证据：`examples/docs/package.json` - **Package**（package_manifest）：{ "private": true, "type": "module", "scripts": { "test": "node --test" } } 证据：`examples/docs/sandbox-agents/repo/package.json` - **Package**（package_manifest）：{ "private": true, "name": "financial-research-agent", "dependencies": { "@openai/agents": "workspace: ", "zod": "^4.0.0" }, "scripts": { "build-check": "tsc --noEmit", "start": "tsx main.ts" } } 证据：`examples/financial-research-agent/package.json` - **Package**（package_manifest）：{ "private": true, "name": "handoffs", "dependencies": { "@openai/agents": "workspace: ", "@openai/agents-core": "workspace: ", "zod": "^4.0.0" }, "scripts": { "build-check": "tsc --noEmit", "start": "tsx index.ts", "start:types": "tsx types.ts", "start:is-enabled": "tsx is-enabled.ts" } } 证据：`examples/handoffs/package.json` - **Package**（package_manifest）：{ "private": true, "name": "mcp", "dependencies": { "@modelcontextprotocol/sdk": "^1.26.0", "@modelcontextprotocol/server-filesystem": "^2026.1.14", "@openai/agents": "workspace: " }, "scripts": { "build-check": "tsc --noEmit", "start:stdio": "tsx filesystem-example.ts", "start:streamable-http": "tsx streamable-http-example.ts", "start:hosted-mcp-on-approval": "tsx hosted-mcp-on-approval.ts", "start:hosted-mcp-human-in-the-loop": "tsx hosted-mcp-human-in-the-loop.ts", "start:hosted-mcp-simple": "tsx hosted-mcp-simple.ts", "start:tool-filter": "tsx tool-filter-example.ts", "start:sse": "tsx sse-example.ts", "start:get-all-tools": "tsx get-all-mcp-tools-example.ts", "start:mcp-servers": "tsx… 证据：`examples/mcp/package.json` - **Package**（package_manifest）：{ "private": true, "name": "memory", "dependencies": { "@openai/agents": "workspace: ", "@prisma/client": "^6.18.0", "zod": "^4.0.0" }, "scripts": { "build-check": "tsc --noEmit", "start:memory": "tsx memory.ts", "start:memory-hitl": "tsx memory-hitl.ts", "start:hitl-session-scenario": "tsx hitl-session-scenario.ts", "start:oai": "tsx oai.ts", "start:oai-hitl": "tsx oai-hitl.ts", "start:oai-compact": "tsx oai-compact.ts", "start:oai-compact-stateless": "tsx oai-compact-stateless.ts", "start:file": "tsx file.ts", "start:file-hitl": "tsx file-hitl.ts", "start:prisma": "tsx start-prisma.ts" }, "devDependencies": { "prisma": "^6.18.0" } } 证据：`examples/memory/package.json` - **Package**（package_manifest）：{ "private": true, "name": "model-providers", "dependencies": { "@openai/agents": "workspace: ", "openai": "^6.35.0", "zod": "^4.0.0" }, "scripts": { "build-check": "tsc --noEmit", "start:custom-example-agent": "tsx custom-example-agent.ts", "start:custom-example-global": "tsx custom-example-global.ts", "start:custom-example-provider": "tsx custom-example-provider.ts" } } 证据：`examples/model-providers/package.json` - **Package**（package_manifest）：{ "name": "nextjs", "private": true, "scripts": { "dev": "next dev --turbopack", "build": "next build", "start": "next start", "lint": "next lint", "build-check": "tsc --noEmit" }, "dependencies": { "@openai/agents": "workspace: ", "@radix-ui/react-dialog": "^1.1.14", "@radix-ui/react-slot": "^1.2.3", "@tanstack/react-query": "^5.80.7", "class-variance-authority": "^0.7.1", "clsx": "^2.1.1", "lucide-react": "^0.515.0", "next": "16.2.3", "react": "^19.0.0", "react-dom": "^19.0.0", "tailwind-merge": "^3.3.1", "wavtools": "^0.1.5", "zod": "^4.0.0" }, "devDependencies": { "@tailwindcss/postcss": "^4", "@types/node": "^20", "@types/react": "^19", "@types/react-dom": "^19", "tailwindcss": "^4", "… 证据：`examples/nextjs/package.json` - **Package**（package_manifest）：{ "name": "realtime-demo", "private": true, "type": "module", "scripts": { "dev": "vite", "build-check": "tsc --noEmit", "build": "tsc && vite build", "preview": "vite preview", "generate-token": "tsx token.ts" }, "devDependencies": { "typescript": "~5.8.3", "vite": "^6.4.2" }, "dependencies": { "@openai/agents-realtime": "workspace: ", "@tailwindcss/vite": "^4.1.7", "openai": "^6.35.0", "tailwindcss": "^4.1.7", "zod": "^4.0.0" } } 证据：`examples/realtime-demo/package.json` - **Package**（package_manifest）：{ "name": "realtime-next", "private": true, "scripts": { "dev": "next dev --turbopack", "build": "next build", "start": "next start", "lint": "next lint", "build-check": "tsc --noEmit" }, "dependencies": { "@openai/agents": "workspace: ", "@radix-ui/react-slot": "^1.2.3", "class-variance-authority": "^0.7.1", "clsx": "^2.1.1", "next": "16.2.3", "react": "^19.0.0", "react-dom": "^19.0.0", "tailwind-merge": "^3.3.0", "wavtools": "^0.1.5", "zod": "^4.0.0" }, "devDependencies": { "@tailwindcss/postcss": "^4", "@types/node": "^20", "@types/react": "^19", "@types/react-dom": "^19", "tailwindcss": "^4", "typescript": "^5" } } 证据：`examples/realtime-next/package.json` - **Package**（package_manifest）：{ "private": true, "name": "realtime-twilio-sip", "dependencies": { "@openai/agents": "workspace: ", "@openai/agents-core": "workspace: ", "@openai/agents-realtime": "workspace: ", "dotenv": "^16.5.0", "fastify": "^5.8.5", "fastify-raw-body": "^5.0.0", "openai": "^6.35.0", "zod": "^4.0.0" }, "scripts": { "build-check": "tsc --noEmit", "start": "tsx server.ts" } } 证据：`examples/realtime-twilio-sip/package.json` - **Package**（package_manifest）：{ "private": true, "name": "realtime-twilio", "dependencies": { "@fastify/formbody": "^8.0.2", "@fastify/websocket": "^11.1.0", "@openai/agents": "workspace: ", "@openai/agents-extensions": "workspace: ", "dotenv": "^16.5.0", "fastify": "^5.8.5", "ws": "^8.18.1", "zod": "^4.0.0" }, "scripts": { "build-check": "tsc --noEmit", "start": "tsx index.ts" } } 证据：`examples/realtime-twilio/package.json` - **Package**（package_manifest）：{ "private": true, "name": "research-bot", "dependencies": { "@openai/agents": "workspace: ", "zod": "^4.0.0" }, "scripts": { "build-check": "tsc --noEmit", "start": "tsx main.ts" } } 证据：`examples/research-bot/package.json` - **Package**（package_manifest）：{ "private": true, "name": "sandbox", "type": "module", "dependencies": { "@openai/agents": "workspace: ", "@openai/agents-extensions": "workspace: ", "zod": "^4.0.0" }, "scripts": { "build-check": "tsc --noEmit", "start:basic": "tsx basic.ts", "start:handoffs": "tsx handoffs.ts", "start:coding-task": "tsx coding-task.ts", "start:resume": "tsx resume.ts", "start:memory": "tsx memory.ts", "start:unix-local-pty": "tsx unix-local-pty.ts", "start:unix-local-runner": "tsx unix-local-runner.ts", "start:docker-runner": "tsx docker-runner.ts", "start:memory-generation": "tsx memory-generation.ts", "start:memory-multi-agent-multiturn": "tsx memory-multi-agent-multiturn.ts", "start:sandbox-agent-capa… 证据：`examples/sandbox/package.json` - **Package**（package_manifest）：{ "private": true, "name": "tools", "type": "module", "dependencies": { "@openai/agents": "workspace: ", "@openai/agents-core": "workspace: ", "@openai/agents-extensions": "workspace: ", "@openai/codex-sdk": "^0.86.0", "chalk": "^5.6.2", "openai": "^6.35.0", "playwright": "^1.55.1", "zod": "^4.0.0" }, "scripts": { "build-check": "tsc --noEmit", "start:codex": "tsx codex.ts", "start:codex-same-thread": "tsx codex-same-thread.ts", "start:computer-use": "tsx computer-use.ts", "start:computer-use-hitl": "tsx computer-use-hitl.ts", "start:file-search": "tsx file-search.ts", "start:tool-search": "tsx tool-search.ts", "start:web-search": "tsx web-search.ts", "start:web-search-filters": "tsx web-se… 证据：`examples/tools/package.json` - **Package**（package_manifest）：{ "name": "bun", "module": "index.ts", "type": "module", "private": true, "scripts": { "start": "bun run index.ts", "start:aisdk": "bun run aisdk.ts", "start:zod": "bun run zod.ts", "start:sandbox:unix-local": "bun run sandbox-unix-local.ts" }, "devDependencies": { "@types/bun": "latest" }, "peerDependencies": { "typescript": "^5" }, "dependencies": { "@openai/agents": "latest", "@openai/agents-extensions": "latest", "zod": "^4" } } 证据：`integration-tests/bun/package.json` - **Package**（package_manifest）：{ "name": "worker", "version": "0.0.0", "private": true, "scripts": { "deploy": "wrangler deploy", "dev": "wrangler dev", "start": "wrangler dev", "test": "vitest", "cf-typegen": "wrangler types", "build": "wrangler deploy --dry-run --outdir dist" }, "dependencies": { "@openai/agents": "latest", "@openai/agents-extensions": "latest" }, "devDependencies": { "typescript": "^5.5.2", "wrangler": "^4.18.0" } } 证据：`integration-tests/cloudflare-workers/worker/package.json` - **Package**（package_manifest）：{ "name": "deno", "private": true, "scripts": { "start": "deno --allow-net --allow-env main.ts" }, "dependencies": { "@openai/agents": "latest", "@openai/agents-extensions": "latest", "zod": "^4" } } 证据：`integration-tests/deno/package.json` - 其余 26 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`, `examples/docs/README.md`, `examples/docs/sandbox-agents/skills/invoice-total-fixer/SKILL.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`, `examples/docs/README.md`, `examples/docs/sandbox-agents/skills/invoice-total-fixer/SKILL.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。 - **Introduction to OpenAI Agents SDK**：importance `high` - source_paths: README.md, packages/agents-core/src/index.ts, package.json - **Installation and Setup**：importance `high` - source_paths: packages/agents-core/package.json, packages/agents/package.json, docs/src/content/docs/guides/quickstart.mdx - **Examples and Patterns**：importance `high` - source_paths: examples/basic/hello-world.ts, examples/agent-patterns/agents-as-tools.ts, examples/mcp/hosted-mcp-simple.ts - **Creating and Running Agents**：importance `high` - source_paths: packages/agents-core/src/agent.ts, packages/agents-core/src/run.ts, packages/agents-core/src/lifecycle.ts, docs/src/content/docs/guides/agents.mdx - **Tools and Tool Use**：importance `high` - source_paths: packages/agents-core/src/tool.ts, packages/agents-core/src/mcp.ts, packages/agents-core/src/mcpUtil.ts, packages/agents-core/src/runner/toolExecution.ts, docs/src/content/docs/guides/tools.mdx - **Guardrails and Input/Output Validation**：importance `medium` - source_paths: packages/agents-core/src/guardrail.ts, packages/agents-core/src/toolGuardrail.ts, packages/agents-core/src/runner/guardrails.ts, docs/src/content/docs/guides/guardrails.mdx - **Handoffs and Multi-Agent Systems**：importance `medium` - source_paths: packages/agents-core/src/handoff.ts, packages/agents-core/src/extensions/handoffFilters.ts, packages/agents-core/src/extensions/handoffPrompt.ts, docs/src/content/docs/guides/handoffs.mdx - **Sandbox Agents Architecture**：importance `high` - source_paths: packages/agents-core/src/sandbox/agent.ts, packages/agents-core/src/sandbox/client.ts, packages/agents-core/src/sandbox/capabilities/index.ts, packages/agents-core/src/sandbox/runtime/manager.ts, packages/agents-core/src/sandbox/manifest.ts ## Repo Inspection Evidence / 源码检查证据 - repo_clone_verified: true - repo_inspection_verified: true - repo_commit: `629d35af99e1ba80fc968b0d062c070caed0683d` - inspected_files: `pnpm-lock.yaml`, `package.json`, `README.md`, `docs/tsconfig.scripts.json`, `docs/package.json`, `docs/README.md`, `docs/tsconfig.json`, `docs/src/content.config.ts`, `docs/src/plugins/typedoc-frontmatter.js`, `docs/src/scripts/translate.ts`, `docs/src/content/docs/index.mdx`, `docs/src/content/docs/ja/index.mdx`, `docs/src/content/docs/guides/multi-agent.md`, `docs/src/content/docs/guides/config.mdx`, `docs/src/content/docs/guides/troubleshooting.mdx`, `docs/src/content/docs/guides/context.mdx`, `docs/src/content/docs/guides/models.mdx`, `docs/src/content/docs/guides/handoffs.mdx`, `docs/src/content/docs/guides/sandbox-agents.mdx`, `docs/src/content/docs/guides/human-in-the-loop.mdx` 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 `openai-agents-js` 与安装入口 `@openai/agents` 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:993521808 | https://github.com/openai/openai-agents-js | repo=openai-agents-js; install=@openai/agents - Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。 ### Constraint 2: 能力判断依赖假设 - Trigger: README/documentation is current enough for a first validation pass. - Host AI rule: 将假设转成下游验证清单。 - Why it matters: 假设不成立时，用户拿不到承诺的能力。 - Evidence: capability.assumptions | github_repo:993521808 | https://github.com/openai/openai-agents-js | README/documentation is current enough for a first validation pass. - Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。 ### Constraint 3: 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:993521808 | https://github.com/openai/openai-agents-js | 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 4: 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:993521808 | https://github.com/openai/openai-agents-js | no_demo; severity=medium - Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。 ### Constraint 5: 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:993521808 | https://github.com/openai/openai-agents-js | 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: 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:993521808 | https://github.com/openai/openai-agents-js | 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 7: 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:993521808 | https://github.com/openai/openai-agents-js | 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。