菜单
Add, refactor, or debug model provider integrations in this repository. Use when the task changes LlmProvider contracts, ProviderCapabilities, OpenAI Responses API handling, streaming, structured outputs, provider routing, or mock-provider parity.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | ai-gateway-provider-integration |
| description | Add, refactor, or debug model provider integrations in this repository. Use when the task changes LlmProvider contracts, ProviderCapabilities, OpenAI Responses API handling, streaming, structured outputs, provider routing, or mock-provider parity. |
Change provider behavior through the provider contract first, then adapt each implementation deliberately.
domain/provider/LlmProvider.java and ProviderCapabilities.java.application/dto.mock and openai behavior aligned enough that service tests can exercise the same contract.ProviderRouter and admin capability exposure if support changes.Responses API orientation in RestClientOpenAiApiClient.LlmProvider../gradlew test --tests '*OpenAiLlmProviderTest' --tests '*ProviderRouterTest'./gradlew test --tests '*AiGatewayServiceTest' --tests '*AiChatControllerTest'AdminControllerTest.references/provider-map.md for the main adapter files and expected parity points.Add or extend end-to-end gateway features in this Spring Boot repository. Use when the task changes request or response DTOs, controller behavior, orchestration in AiGatewayService, admin APIs, quota or audit behavior, or any cross-layer feature that must preserve the repository's API, application, domain, and infrastructure layering.
Harden security, guardrails, and policy enforcement in this repository. Use when the task affects API key auth, tenant scope, admin access, rate limiting, actuator exposure, moderation, guardrail rules, error responses, or any fail-closed behavior in the gateway.
Implement or refine tool-call orchestration in this repository. Use when the task adds executable tools, tool request schemas, tool policy enforcement, model-issued tool calls, follow-up responses, or SSE behavior for tool execution loops.