One Declaration Generates a CLI, Agent Skills, and Docs from Company APIs

Teams wiring internal APIs into AI coding agents hit the same wall: authentication, inconsistent output, and brittle error handling. This SDK standardizes those three layers so an Agent gets a predictable contract—fixed JSON shapes, typed exit codes, and token lifecycle management it never touches—without requiring a persistent MCP server.

`@renxqoo/agent-cli-sdk` generates a CLI command, a SKILL.md file for AI Agents, and syncs to Agent directories from one `defineCommand` call. Authentication is handled via OAuth 2.1 plugins that keep tokens on disk and away from the Agent, refreshing them silently and surfacing deterministic exit codes when re-login is needed. The SDK enforces a fixed JSON output contract for both success and failure, paired with nine typed error classes and a detailed exit-code table so Agents can branch reliably without parsing unpredictable text. A built-in Skill Factory lets an Agent itself generate the CLI wrapper from an API description, collapsing the write-command, write-docs, and Agent-adaptation steps into a single workflow. The project argues CLI is a better fit than MCP for Agent tooling because it avoids persistent server processes, constant context injection, and protocol overhead, while fitting naturally into Unix pipelines and subprocess calls from any language.