[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6AC5S48FHHDSWTY69F07P9Z":3,"skill-same:01M6AC5S48FHHDSWTY69F07P9Z":55},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":21,"linkOnly":22,"markdownContent":23,"markdownBody":24,"frontMatter":25,"refName":26,"commitSha":27,"skillDirectory":28,"forksCount":29,"repositoryHtmlUrl":30,"repositoryHomepageUrl":31,"ownerHtmlUrl":32,"ownerBlogUrl":33,"ownerDescription":6,"updatedAt":34,"install":35,"resources":48},"01M6AC5S48FHHDSWTY69F07P9Z","implementation-strategy",null,"Choose bounded implementation scope and existing owning modules before Guardrails runtime, configuration, client, resource, streaming, Agents, or SDK compatibility changes and feedback fixes.","openai\u002Fopenai-guardrails-js","openai","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F14957082?v=4","01M6AC29YGCR4S2NTJFV0G8B81","OpenAI","01M6AC5JPN01DTSJMFDT21GDMJ",113,"MIT","https:\u002F\u002Fgithub.com\u002Fopenai\u002Fopenai-guardrails-js\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fimplementation-strategy\u002FSKILL.md",[],[19,20],"implementation strategy","sdk","2026-11-10 10:12:30",false,"---\nname: implementation-strategy\ndescription: Choose bounded implementation scope and existing owning modules before Guardrails runtime, configuration, client, resource, streaming, Agents, or SDK compatibility changes and feedback fixes.\n---\n\n# Implementation Strategy\n\nBefore implementation or a feedback fix, record the requested outcome,\nacceptance criteria, affected paths, compatibility requirements, and non-goals.\nApply [scope discipline](..\u002F..\u002F..\u002FAGENTS.md#task-scope-and-review-discipline).\nAn issue's suggested implementation is evidence, not an approved API design.\n\n## Find the owning boundary\n\nRead the [repository map](..\u002F..\u002F..\u002FAGENTS.md#repository-map-and-supported-tools)\nand [SDK migration guide](..\u002F..\u002F..\u002Fdocs\u002Fsdk_migration.md). Use existing\nconfiguration, validation, registry, conversation, and resource pipelines.\nIdentify whether the affected contract is a public export, declaration,\nconfiguration format, request\u002Fresponse shape, exception, CLI option, or internal\nhelper. Compare ownership changes against the intended PR base; consult a\nverified release tag separately when evaluating released compatibility. Report\nan unavailable or stale release baseline rather than treating branch-only\nbehavior as a released guarantee.\n\n| Boundary | Preserve and inspect |\n| --- | --- |\n| OpenAI\u002FAzure clients | Async create factories, separate check client, caller options and credentials, and initialized guarded clones from `withOptions()` |\n| Resources | Chat\u002FResponses parameters and request options, SDK-compatible types, and the explicit `guardrails` namespace; do not imply every inherited SDK method is guarded |\n| OpenAI 7 | Native Fetch `Response`\u002F`Headers`, provider-specific options, and inherited SDK method contracts |\n| Zod 4 | Existing schemas and their owning validation boundary; `GuardrailSpec.schema()` returns the definition, not JSON Schema |\n| Agents | Public `@openai\u002Fagents` imports, built CommonJS behavior, session history, and preflight guards blocking model dispatch |\n| Pipeline | `pre_flight`, `input`, `output` ordering and each integration's documented execution model; distinguish tripwires from execution errors |\n| Streaming and history | Resource-specific output checks, conversation ordering and roles, tool-call identity, masking, and output error policy |\n\nDo not add new guarded SDK methods, normalize omitted options into arbitrary\ndefaults, or duplicate schema\u002Fconversation conversion just to accommodate a\nsuggested implementation. Escalate a substantive public API or architecture\nchange before expanding the diff.\n\n## Match tests to the changed contract\n\nUse existing unit tests beside the owning area. For SDK boundaries, inspect\n`src\u002F__tests__\u002Fintegration\u002Fsdk-compat.test.ts`, `sdk-types.test.ts`, and\n`sdk-migration-feedback.test.ts`: they cover built-package behavior and consumer\ntypes that source-only mocks can miss. Build before running these tests. Use\nVitest mocks\u002Fspies or SDK-provided testing utilities where they model the real\nprotocol. Add only tests needed for the changed behavior.\n\nFor a touched security surface, review the relevant trust boundary: untrusted\nconfiguration, provider responses, conversation content, or eval inputs, and\nwhere validation occurs before side effects. Keep this review scoped to the\nchange; do not convert it into unrelated hardening.\n\nClassify feedback using AGENTS.md. Fix introduced\u002Fworsened defects and narrowly\nnecessary corrections. Record broader improvements separately. If repeated\nspecial cases suggest the approach fights an existing invariant, revisit the\nsmallest design at the owning boundary before adding more machinery.\n\nUse [code-change-verification](..\u002Fcode-change-verification\u002FSKILL.md) for checks\nand the [release guide](..\u002F..\u002F..\u002F.changeset\u002FREADME.md) for changeset impact.\n","# Implementation Strategy\n\nBefore implementation or a feedback fix, record the requested outcome,\nacceptance criteria, affected paths, compatibility requirements, and non-goals.\nApply [scope discipline](..\u002F..\u002F..\u002FAGENTS.md#task-scope-and-review-discipline).\nAn issue's suggested implementation is evidence, not an approved API design.\n\n## Find the owning boundary\n\nRead the [repository map](..\u002F..\u002F..\u002FAGENTS.md#repository-map-and-supported-tools)\nand [SDK migration guide](..\u002F..\u002F..\u002Fdocs\u002Fsdk_migration.md). Use existing\nconfiguration, validation, registry, conversation, and resource pipelines.\nIdentify whether the affected contract is a public export, declaration,\nconfiguration format, request\u002Fresponse shape, exception, CLI option, or internal\nhelper. Compare ownership changes against the intended PR base; consult a\nverified release tag separately when evaluating released compatibility. Report\nan unavailable or stale release baseline rather than treating branch-only\nbehavior as a released guarantee.\n\n| Boundary | Preserve and inspect |\n| --- | --- |\n| OpenAI\u002FAzure clients | Async create factories, separate check client, caller options and credentials, and initialized guarded clones from `withOptions()` |\n| Resources | Chat\u002FResponses parameters and request options, SDK-compatible types, and the explicit `guardrails` namespace; do not imply every inherited SDK method is guarded |\n| OpenAI 7 | Native Fetch `Response`\u002F`Headers`, provider-specific options, and inherited SDK method contracts |\n| Zod 4 | Existing schemas and their owning validation boundary; `GuardrailSpec.schema()` returns the definition, not JSON Schema |\n| Agents | Public `@openai\u002Fagents` imports, built CommonJS behavior, session history, and preflight guards blocking model dispatch |\n| Pipeline | `pre_flight`, `input`, `output` ordering and each integration's documented execution model; distinguish tripwires from execution errors |\n| Streaming and history | Resource-specific output checks, conversation ordering and roles, tool-call identity, masking, and output error policy |\n\nDo not add new guarded SDK methods, normalize omitted options into arbitrary\ndefaults, or duplicate schema\u002Fconversation conversion just to accommodate a\nsuggested implementation. Escalate a substantive public API or architecture\nchange before expanding the diff.\n\n## Match tests to the changed contract\n\nUse existing unit tests beside the owning area. For SDK boundaries, inspect\n`src\u002F__tests__\u002Fintegration\u002Fsdk-compat.test.ts`, `sdk-types.test.ts`, and\n`sdk-migration-feedback.test.ts`: they cover built-package behavior and consumer\ntypes that source-only mocks can miss. Build before running these tests. Use\nVitest mocks\u002Fspies or SDK-provided testing utilities where they model the real\nprotocol. Add only tests needed for the changed behavior.\n\nFor a touched security surface, review the relevant trust boundary: untrusted\nconfiguration, provider responses, conversation content, or eval inputs, and\nwhere validation occurs before side effects. Keep this review scoped to the\nchange; do not convert it into unrelated hardening.\n\nClassify feedback using AGENTS.md. Fix introduced\u002Fworsened defects and narrowly\nnecessary corrections. Record broader improvements separately. If repeated\nspecial cases suggest the approach fights an existing invariant, revisit the\nsmallest design at the owning boundary before adding more machinery.\n\nUse [code-change-verification](..\u002Fcode-change-verification\u002FSKILL.md) for checks\nand the [release guide](..\u002F..\u002F..\u002F.changeset\u002FREADME.md) for changeset impact.\n",{"name":5,"description":7},"main","1f5b2d5d31f70e4766800da41d23aefdf1fbdf04",".agents\u002Fskills\u002Fimplementation-strategy",22,"https:\u002F\u002Fgithub.com\u002Fopenai\u002Fopenai-guardrails-js","https:\u002F\u002Fguardrails.openai.com","https:\u002F\u002Fgithub.com\u002Fopenai","https:\u002F\u002Fopenai.com\u002F","2026-11-11 05:23:00",{"source":8,"skillName":5,"commands":36,"prompt":46,"zipAvailable":47},[37,40,43],{"tool":38,"command":39},"npx","npx skills add openai\u002Fopenai-guardrails-js --skill implementation-strategy",{"tool":41,"command":42},"bunx","bunx skills add openai\u002Fopenai-guardrails-js --skill implementation-strategy",{"tool":44,"command":45},"pnpm","pnpm dlx skills add openai\u002Fopenai-guardrails-js --skill implementation-strategy","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fopenai\u002Fopenai-guardrails-js\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fimplementation-strategy\u002FSKILL.md",true,[49],{"path":50,"resourceType":51,"resourceTypeDesc":52,"sizeBytes":53,"sourceHtmlUrl":54},"agents\u002Fopenai.yaml","REFERENCE","参考文档",127,"https:\u002F\u002Fgithub.com\u002Fopenai\u002Fopenai-guardrails-js\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fimplementation-strategy\u002Fagents\u002Fopenai.yaml",[56,66,75,85],{"skillNo":57,"name":58,"nameZh":6,"description":59,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":60,"categories":61,"keywords":62,"publishedAt":21},"01M6AC5T7FZE6EDYK5PBXQKADB","pr-draft-summary","Draft a Guardrails PR from its complete final diff and carry out authorized CI and review follow-up, including stacked PRs and takeovers.","https:\u002F\u002Fgithub.com\u002Fopenai\u002Fopenai-guardrails-js\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fpr-draft-summary\u002FSKILL.md",[],[63,64,65],"pr draft summary","pr","review",{"skillNo":67,"name":68,"nameZh":6,"description":69,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":70,"categories":71,"keywords":72,"publishedAt":21},"01M6AC5QAF7ZST45DS0PEKW9JK","implementation-kickoff","Start or resume a requested Guardrails implementation or PR takeover in the selected linked worktree with bounded scope and verification.","https:\u002F\u002Fgithub.com\u002Fopenai\u002Fopenai-guardrails-js\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fimplementation-kickoff\u002FSKILL.md",[],[73,74,64],"implementation kickoff","verification",{"skillNo":76,"name":77,"nameZh":6,"description":78,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":79,"categories":80,"keywords":81,"publishedAt":21},"01M6AC5NWMTSRH2B62H8Z1QX8G","implementation-final-review","Prepare independent review of a complete Guardrails change before pushing or updating a PR, using the repository adversarial-review procedure.","https:\u002F\u002Fgithub.com\u002Fopenai\u002Fopenai-guardrails-js\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fimplementation-final-review\u002FSKILL.md",[],[65,82,83,64,84],"implementation final review","change","prepare",{"skillNo":86,"name":87,"nameZh":6,"description":88,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":89,"categories":90,"keywords":96,"publishedAt":21},"01M6AC5MB35097EGZ1RPNZZWFK","code-change-verification","Select and run Guardrails verification for code, dependency, test, tooling, documentation, or repository-workflow changes.","https:\u002F\u002Fgithub.com\u002Fopenai\u002Fopenai-guardrails-js\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fcode-change-verification\u002FSKILL.md",[91],{"categoryNo":92,"categoryName":93,"taxonomyType":94,"taxonomyTypeDesc":95},"01JBSEEDCAT000000000000030","开发工程师","PROFESSION","适用职业",[74,97,83,98,99,100,101,102],"code","code change verification","workflow","run","test","documentation"]