What problem does it solve? Choosing the wrong SDK packages, APIs, or runtime when building a programmatic agent integration leads to rewrites and broken deployments. This Skill identifies whether the target is a current OpenAI Codex runtime or a legacy Cursor-compatible host before any code is written, then guides the integration accordingly. ## Core Features & Use Cases - Runtime Detection: Distinguishes current Codex integrations from legacy Cursor-compatible hosts so the correct SDK packages, paths, and tool schemas are used. - Language Inference: Infers TypeScript versus Python from repository signals like package.json, pyproject.toml, or existing source files, stating a reversible default when signals are absent. - Legacy Reference Gating: Loads the legacy Cursor SDK reference (covering Agent.prompt, Agent.create, Agent.resume, local vs cloud runtimes, MCP servers, and disposal patterns) only after the host is explicitly verified as Cursor-compatible. - Use Case: A developer asks to automate code review agents in CI. The Skill checks the runtime, picks the correct SDK and language, and produces a working integration without mixing Codex and Cursor APIs. ## Quick Start Use the projectb-sdk skill to build a programmatic agent integration for this repository and tell me which runtime and SDK packages to use.