RichNasz

Use when given a Google Doc spec and a destination Google Doc to write a generated article into — reads all tabs from the source doc sequentially as a prompt chain, generates the article, then writes it to a named tab in the destination doc

Reads a Spec Coach report from a generated article doc and automatically applies the recommended improvements to the spec document. Accepts a YAML config file that identifies the spec doc, article doc, and any number of reference docs for research-backed improvements. Reports all changes made, including tab reorders performed via the create-verify-delete pattern.

Use after generate-article to evaluate how well a multi-tab Google Doc spec performed — scores the generated article against the spec's own criteria, analyzes semantic drift across the tab chain, and writes improvement recommendations to a "Spec Coach" tab in the article document

Shared utility scripts for Google Workspace document operations via the gws CLI. Required by generate-article, spec-coach, and spec-auto-tune. Not invoked directly.

Helps the agent use SwiftOpenResponsesDSL to build type-safe LLM request pipelines for the Open Responses API with tool calling, sessions, and persistent agents with conversation continuity via previous_response_id. Useful when defining ResponseRequest objects, wiring LLMTool instances into a ToolSession or Agent, chaining responses across turns, or processing streaming events in Swift.

Step-by-step process for designing and implementing a Swift application that uses SwiftOpenResponsesDSL for the Open Responses API. Use this when translating requirements into a working LLM integration with tool calling, streaming, or multi-turn conversations. Complements the using-swift-open-responses-dsl reference skill. Scoped to DSL wiring only -- does not cover @LLMTool struct design.

Helps the agent use SwiftChatCompletionsDSL to build type-safe LLM request pipelines with tool calling, sessions, and persistent agents for any OpenAI-compatible API. Useful when defining chat completions requests, wiring LLMTool instances into a ToolSession or Agent, handling multi-turn conversations, or processing streaming responses in Swift.