cortex-reference

Answers conceptual and functional questions about the Cortex conversational agent platform.

Updated Sep 17, 2026
One-click install
npx skills add https://github.com/karenrebecag/spec-driven-standards --skill cortex-reference-karenrebecag
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: cortex-reference
Source: https://github.com/karenrebecag/spec-driven-standards/tree/main/plugins/atom/skills/cortex-reference
Command: npx skills add https://github.com/karenrebecag/spec-driven-standards --skill cortex-reference-karenrebecag

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Teams building conversational agents on Atom's Cortex platform need accurate answers about what each panel, node, data type, and feature does without digging through a 46,000-word manual, and this Skill routes each question to the right reference section. ## Core Features & Use Cases - Conceptual lookup: Explains Cortex terms and features such as nodes, funnel stages, tipificaciones, etiquetas, condition branches, save fields, RAG knowledge bases, dynamic tables, HTTP Request, Code Tool, MCP Servers, OAuth apps, WhatsApp Flows, the simulator, evaluations, and draft vs published states. - Organized reference manual: Routes questions to 10 section files under references/ covering fundamentals, global configuration, flow building, knowledge and data, tools and integrations, messaging, testing, publishing, Flowbuilder integration, and best practices. - Draft-status awareness: Flags that 39 of 41 source articles are marked as drafts, so unverified operational claims are confirmed against the live product before acting. - Use Case: A builder asks how a dynamic table differs from a knowledge base for pricing data; the Skill reads references/04-conocimiento-y-datos.md and explains when to use each. ## Quick Start Ask what a Cortex term means or how a specific platform feature is configured, for example how condition branches route conversations between nodes.

Frequently Asked Questions about cortex-reference

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
What is Cortex and when should I use it instead of Flowbuilder?

Cortex is Atom's conversational agent builder for conversations that cannot be scripted in advance, where customers write freely and the AI must interpret intent. Flowbuilder fits linear, predictable flows with buttons; Cortex fits support, lead qualification, scheduling, and quoting.

When should I use a dynamic table versus a knowledge base in Cortex?

Use a knowledge base (RAG over documents) for unstructured information that rarely changes, like manuals and FAQs. Use a dynamic table connected to an API or Google Sheets for structured data that changes often, like stock, prices, or availability.

How do condition branches route conversations between Cortex nodes?

Each branch between nodes carries a natural-language condition that the LLM evaluates at runtime to decide the transfer. Conditions should be mutually exclusive and concrete; pairs exceeding 80% semantic similarity are flagged as conflicts.

Is the Cortex reference documentation fully validated?

No. Thirty-nine of the forty-one source articles are marked as drafts and only two are in review, so the content is the best available description rather than confirmed truth. For irreversible or production-affecting actions, verify against the live product or the Cortex team.

What is the difference between cortex-reference and cortex-implementation?

cortex-reference is conceptual, explaining what each Cortex piece is and how it is used in the UI. cortex-implementation is operational, guiding the full process of migrating an existing flow or building an agent from scratch with checklists and known platform bugs.