Skill Library.
Reusable agent behaviors. Some are universal across every Onyx and Pynk deployment; some are industry-specific; some were built for a specific customer and have been generalized back to the catalog.
after-hours-emergency
uncategorized
After hours, an HVAC operator's phone gets two kinds of messages: "the AC is out and my mom is on oxygen" and "want to schedule a tune-up next week." The first is a real emergency that needs the opera
AmazonBedrock
reference
This course is intended to provide you with a comprehensive step-by-step understanding of how to engineer optimal prompts within Claude, using Bedrock.
anthropic_api_fundamentals
reference
A series of notebook tutorials that cover the essentials of working with Claude models and the Anthropic SDK including:
bootstrap-realtime-eval
reference
Bootstrap a new realtime eval folder inside this cookbook repo by choosing the right harness from examples/evals/realtime_evals, scaffolding prompt/tools/data files, generating a useful README, and validating it with smoke, full eval, and test runs. Use when a user wants to start a new crawl, walk, or run realtime eval in this repository.
browser-translation-demo
reference
This is a small browser demo for one-way live translation from tab audio. The
building-a-supply-chain-copilot-with-agent-sdk-and-databricks-mcp
reference
A full-stack, Databricks-themed conversational assistant for supply chain queries, powered by OpenAI Agents and Databricks MCP servers. Includes a React chat UI and a FastAPI backend that streams agent responses.
case_summary_service
reference
Small offline fixture for the sandboxed migration cookbook.
cassandra_astradb
reference
The demos in this directory show how to use the Vector
chatgpt_prompt_guide
reference
This guide is for Enterprise users who want more reliable results from ChatGPT without needing advanced technical expertise. It is a practical, example-driven guide that is easy to apply to your everyday work, whether that’s summarizing documents, drafting emails, translating content, analyzing simple scenarios, or brainstorming narratives.
codex
reference
When deploying production code, most teams rely on CI/CD pipelines to validate changes before merging. Reviewers typically look at unit test results, vulnerability scans, and code quality reports. Traditionally, these are produced by rule-based engines that catch known issues but often miss contextual or higher-order problems—while leaving developers with noisy results that are hard to prioritize
comfyui
uncategorized
"Generate images, video, and audio with ComfyUI — install, launch, manage nodes/models, run workflows with parameter injection. Uses the official comfy-cli for lifecycle and direct REST/WebSocket API
crawl_harness
reference
Deterministic single-turn replay for Realtime evals. This harness feeds fixed audio
customer-objection-handling
uncategorized
Equips the customer's Onyx agent to handle the recurring objections their callers/messagers raise — price, timing, trust, scope, comparison — without lapsing into salesy script tone or making promises
deployment_manager
reference
Local control-plane app for running and observing Agents SDK demo projects.
domain-intel
uncategorized
Passive domain reconnaissance using Python stdlib. Use this skill for subdomain discovery, SSL certificate inspection, WHOIS lookups, DNS records, domain availability checks, and bulk multi-domain ana
editing_harness
reference
**Best for:** high-precision checks where edits must be exact and non-target
elasticsearch
reference
Elasticsearch is a popular search/analytics engine and [vector database](https://www.elastic.co/elasticsearch/vector-database).
escalation-to-human
uncategorized
The agent's pressure-release valve. Defines the conditions under which a customer agent must stop responding and route the conversation to a human (the operator) — with the right framing so the custom
evaluation
reference
This cookbook provides a practical guide on how to use the OpenAI Platform to easily build resilience into your prompts.
everything
mcp
**[Architecture](docs/architecture.md)
fetch
mcp
<!-- mcp-name: io.github.modelcontextprotocol/server-fetch -->
filesystem
mcp
Node.js server implementing Model Context Protocol (MCP) for filesystem operations.
generation_harness
reference
**Best for:** fast iteration on prompt-following, layout, and text-rendering
git
mcp
<!-- mcp-name: io.github.modelcontextprotocol/server-git -->
gpt_actions_library
reference
This page provides an instruction & guide for developers building middleware to connect a GPT Action to a specific application. Before you proceed, make sure to first familiarize yourself with the following information:
hermes-agent
uncategorized
Complete guide to using and extending Hermes Agent — CLI usage, setup, configuration, spawning additional agents, gateway platforms, skills, voice, tools, profiles, and a concise contributor reference
Hospitality Guest Recovery
industry
When something goes wrong on a stay — a missed booking, a billing error, a service slip — the agent recognizes it, acknowledges it in the property's voice, and takes the recovery action the property is prepared to authorize.
how_to_build_a_deep_research_mcp_server
reference
This is a minimal example of a Deep Research style MCP server for searching and fetching files from the OpenAI file storage service.
imagegen_evals
reference
This folder contains a lightweight vision eval harness plus example runners for
kanban-orchestrator
uncategorized
Decomposition playbook + specialist-roster conventions + anti-temptation rules for an orchestrator profile routing work through Kanban. The "don't do the work yourself" rule and the basic lifecycle ar
kanban-worker
uncategorized
Pitfalls, examples, and edge cases for Hermes Kanban workers. The lifecycle itself is auto-injected into every worker's system prompt as KANBAN_GUIDANCE (from agent/prompt_builder.py); this skill is w
kusto
reference
[Azure Data Explorer aka Kusto](https://azure.microsoft.com/en-us/products/data-explorer) is a cloud-based data analytics service that enables users to perform advanced analytics on large datasets in real-time. It is particularly well-suited for handling large volumes of data, making it an excellent choice for storing and searching vectors.
lead-qualification
uncategorized
A real-estate agent gets a lead from Zillow at 2pm and another from their IDX form at 2:04pm. By 2:15pm both leads have already messaged the next agent on the search results, because the first agent w
listing-followup
uncategorized
After a showing or open house, the right follow-up moment is somewhere between "too soon" (annoying — the buyer is still driving home) and "too late" (the lead is cold and the buyer has already toured
livekit-translation-demo
reference
Local Next.js 16 demo for a LiveKit video room with per-listener OpenAI
Meeting Summarization
universal
Turns a call or meeting recording into a structured summary — decisions, action items, open questions, deferred topics — written in the business's voice and addressed to the right people.
memory
mcp
A basic implementation of persistent memory using a local knowledge graph. This lets Claude remember information about the user across chats.
migration_agent
reference
You are migrating the mounted repo under `repo/`.
missed-call-response
uncategorized
Home-services operators bleed revenue on missed calls. Plumber is under a sink, HVAC tech is on a roof, owner-operator is on another job — phone rings and rolls to voicemail. Most callers don't leave
ml-paper-writing
uncategorized
Write publication-ready ML/AI papers for NeurIPS, ICML, ICLR, ACL, AAAI, COLM. Use when drafting papers from research repos, structuring arguments, verifying citations, or preparing camera-ready submi
mongodb_atlas
reference
[Atlas Vector Search](https://www.mongodb.com/products/platform/atlas-vector-search) is a fully managed service that simplifies the process of effectively indexing high-dimensional vector data within MongoDB and being able to perform fast vector similarity searches. With Atlas Vector Search, you can use MongoDB as a standalone vector database for a new project or augment your existing MongoDB coll
Multilingual Outreach
universal
Detects the language a customer prefers and responds in it — natively, not via a visible 'translated by' banner — while keeping the business's voice intact across languages.
neon
reference
[Neon](https://neon.tech/) is Serverless Postgres built for the cloud. Neon separates compute and storage to offer modern developer features such as autoscaling, database branching, scale-to-zero, and more.
one_way_translation_using_realtime_api
reference
This project demonstrates how to use the [OpenAI Realtime API](https://platform.openai.com/docs/guides/realtime) to build a one-way translation application with WebSockets. It is implemented using the [Realtime + Websockets integration](https://platform.openai.com/docs/guides/realtime-websocket). A real-world use case for this demo is multilingual, conversational translation—where a speaker talks
open-house-confirmations
uncategorized
A real-estate agent runs an open house Sunday 1-3pm and 14 people RSVP'd through Zillow, the IDX form, and the operator's website. By Sunday morning, half forgot they RSVP'd. Of the half that show, a
overnight-mission-watch
uncategorized
Set up an overnight SENTINEL-style watch for an active coding mission using cron jobs plus a pre-run snapshot script. Covers heartbeat, spend, stop-criteria escalation, and scheduled morning brief del
pinecone
reference
[Vector search](https://www.pinecone.io/learn/vector-search-basics/) is an innovative technology that enables developers and engineers to efficiently store, search, and recommend information by representing complex data as mathematical vectors. By comparing the similarities between these vectors, you can quickly retrieve relevant information in a seamless and intuitive manner.
pretext
uncategorized
"Use when building creative browser demos with @chenglou/pretext — DOM-free text layout for ASCII art, typographic flow around obstacles, text-as-geometry games, kinetic typography, and text-powered g
product-execution
uncategorized
"You are the execution engine for two products: PynkBrain (AI social network) and OnyxBrain (AI employee for small business). Both products have Gemini-powered chat interfaces. You handle the actions
Professional Services Intake
industry
Handles new-client intake for a professional services firm — legal, accounting, advisory — gathering matter details, conflicts, scope, and engagement constraints in the firm's voice and writing the result back to the practice management system.
prompt_engineering_interactive_tutorial
reference
This course is intended to provide you with a comprehensive step-by-step understanding of how to engineer optimal prompts within Claude.
prompt_evaluations
reference
Welcome to Anthropic's comprehensive prompt evaluations course. Across nine lessons, you will learn everything you need to know to implement evaluations successfully in your workflows with the Anthropic API. We recommend that you start from the beginning with the [Evaluations 101](./01_intro_to_evals/01_intro_to_evals.ipynb) lesson, as each lesson builds on key concepts taught in previous ones.
prompts
reference
If a tool call fails due to an authentication or server error (such as a 500 Internal Server Error, or 4XX errors), timeout, or network issue, you MUST retry the same tool call up to 2 more times before giving up. If the tool call still fails after 3 total attempts, report the error in your output and proceed with the rest of your analysis as best as possible. In situations where there isn't an ex
Real Estate Lead Qualification
industry
Qualifies inbound real estate leads against the brokerage's actual criteria (timeline, financing, location, price band) and routes them to the right agent — instead of leaving every lead in a single dump for the team to fight over.
real_world_prompting
reference
Welcome to Anthropic's comprehensive real world prompting tutorial. This course is designed for experienced developers who have already dipped their toes into the world of prompt engineering, particularly those who have completed our comprehensive **[Prompt engineering interactive tutorial](../prompt_engineering_interactive_tutorial/README.md)**. If you haven't gone through that tutorial yet, we s
realtime_evals
reference
This folder contains three evaluation harnesses for the Realtime API that increase in complexity:
redis
reference
Most developers from a web services background are probably familiar with Redis. At it's core, Redis is an open-source key-value store that can be used as a cache, message broker, and database. Developers choice Redis because it is fast, has a large ecosystem of client libraries, and has been deployed by major enterprises for years.
references
reference
Use this table when the user is not sure which realtime eval harness to start with.
results_viewer
reference
This directory contains a [Streamlit](https://streamlit.io/) app for exploring
run_harness
reference
Model-simulated multi-turn eval harness for the Realtime API. This harness uses
seasonal-promotions
uncategorized
Home-services operators sell things that follow weather: HVAC tune-up before summer, gutter cleaning before fall, drain camera inspection before holiday cooking, garage door weatherstripping before wi
sequentialthinking
mcp
An MCP server implementation that provides a tool for dynamic and reflective problem-solving through a structured thinking process.
service-call-scheduling
uncategorized
Most home-services operators schedule jobs by texting back and forth: "How about Tuesday?" "I work nights." "Wednesday?" "Morning?" Three days and six messages later, a slot is booked — and half of th
sharepoint_azure_function
reference
This solution enables a GPT action to answer a user’s question with the context of files the user can access in SharePoint or Office365, using Microsoft’s Graph API [search capabilities](https://learn.microsoft.com/en-us/graph/api/resources/search-api-overview?view=graph-rest-1.0) and the ability to [retrieve files](https://learn.microsoft.com/en-us/graph/api/driveitem-get?view=graph-rest-1.0\&tab
showing-scheduling
uncategorized
A buyer's agent gets "can we see 47 Maple this weekend?" at 9pm Friday and another "want to see Friday 6pm at 312 Oak" at 10pm. By morning the buyer has booked someone else for both. This skill handle
SingleStoreDB
reference
**[SingleStoreDB](https://singlestore.com)** has first-class support for vector search through our [Vector Functions](https://docs.singlestore.com/managed-service/en/reference/sql-reference/vector-functions.html). Our vector database subsystem, first made available in 2017 and subsequently enhanced, allows extremely fast nearest-neighbor search to find objects that are semantically similar, easily
structured_outputs_articles
reference
Mixture of Experts (MoE) is a machine learning technique designed to enhance model performance by combining the predictions of multiple specialized models, or "experts." The concept was introduced by Michael I. Jordan and Robert A. Jacobs in 1991. MoE models have been applied in various fields, including natural language processing, computer vision, and speech recognition, to improve accuracy and
supabase
reference
[Supabase](https://supabase.com/docs) is an open-source Firebase alternative built on top of [Postgres](https://en.wikipedia.org/wiki/PostgreSQL), a production-grade SQL database.
support_reply_service
reference
This tiny package drafts a support-agent reply with the OpenAI Python client.
third_party
reference
This guide explains how to integrate OpenAI reasoning models into your GitHub Pull Request (PR) workflow to automatically review code for quality, security, and enterprise standards compliance. By leveraging AI-driven insights early in the development process, you can catch issues sooner, reduce manual effort, and maintain consistent best practices across your codebase.
time
mcp
<!-- mcp-name: io.github.modelcontextprotocol/server-time -->
tool_use
reference
Welcome to Anthropic's comprehensive tool use tutorial. Across six lessons, you will learn everything you need to know to implement tool use successfully in your workflows with Claude. We recommend that you start from the beginning with the [tool use overview](./01_tool_use_overview.ipynb), as each lesson builds on key concepts taught in previous ones.
twilio-translation-demo
reference
This is a server-side demo for two-person phone translation with Twilio Media Streams and `gpt-realtime-translate`.
typesense
reference
Typesense is an open source, in-memory search engine, that you can either [self-host](https://typesense.org/docs/guide/install-typesense.html#option-2-local-machine-self-hosting) or run on [Typesense Cloud](https://cloud.typesense.org/).
ui
reference
This template provides a minimal setup to get React working in Vite with HMR and some ESLint rules.
vector_databases
reference
This section of the OpenAI Cookbook showcases many of the vector databases available to support your semantic search use cases.
voice_solutions
reference
voice-pattern-mirroring
uncategorized
Captures the customer's actual speech patterns from discovery — phrasing, sentence rhythm, contractions, what they say, what they refuse to say — and gives the customer's Hermes-CustomerSuccess agent
walk_harness
reference
The walk harness replays saved audio to make realtime eval runs comparable.
weaviate
reference
[Weaviate](https://weaviate.io) is an open-source vector search engine ([docs](https://weaviate.io/developers/weaviate) - [Github](https://github.com/weaviate/weaviate)) that can store and search through OpenAI embeddings _and_ data objects. The database allows you to do similarity search, hybrid search (the combining of multiple search techniques, such as keyword-based and vector search), and ge
xitter
uncategorized
Interact with X/Twitter via the x-cli terminal client using official X API credentials. Use for posting, reading timelines, searching tweets, liking, retweeting, bookmarks, mentions, and user lookups.