Workflow Packs
User Guide
Use Descrybe Legal Research Workflows
Ready-made assistant workflows for source-grounded legal research
This guide explains the public Descrybe Legal Research Workflows repository: what it contains, when to use it, and how it fits with Descrybe Legal Engine and the Python SDK.
Overview
What the workflow repository is for
Descrybe Legal Research Workflows is a public repository of assistant-facing legal research patterns powered by Descrybe Legal Engine.
Use it when you want a repeatable research workflow inside an assistant such as Claude, ChatGPT, Codex, or another compatible surface. The repository is not the legal research service itself. Descrybe Legal Engine provides the source-law tools; the workflow pack helps the assistant use those tools in a more structured way.
A simple way to think about it
- Descrybe Legal Engine is the research source and MCP service.
- The workflow repository is the assistant-facing recipe book.
- The Python SDK is the app-building layer for scripts, agents, and custom tools.
Included workflows
Research Roadmap
Use this when you have a plain-English legal issue and need a research path, not a final answer. The workflow helps frame the issue, identify assumptions, and point the assistant toward source-grounded research tasks.
Authority Finder
Use this when you have a legal proposition and need cases that support, limit, distinguish, or reject it. The workflow is built around finding authority and preserving the source trail for review.
Citation and Quote Auditor
Use this when a draft, memo, argument, or passage has citations and quoted language that need checking against case law.
The repository also includes setup documentation, surface guidance, examples, and packaging for Claude-style and OpenAI-oriented assistant workflows.
Choose a surface
Claude
Use the Claude-oriented workflow pack when you want Descrybe Legal Engine configured as the legal research connector for Claude-style skills and workflows. Start from the workflow repository.
ChatGPT, Codex, or OpenAI-oriented plugins
Use the OpenAI-oriented workflow pack when you want the same research patterns packaged for OpenAI-style plugin or coding-agent workflows. Start from the workflow repository.
Python app or script
Use the Python SDK guide when you are building the application, OAuth flow, token storage, or custom agent code. The workflow repository can still help with prompts and research patterns.
How to start
- 1
Confirm Descrybe access
You need a Descrybe account with Legal Engine or Platform access before the workflows can use the hosted Descrybe service.
- 2
Open the workflow repository
Start with the public repository README, then choose the setup path for the assistant or coding environment you plan to use.
- 3
Install or copy the relevant pack
Use the Claude-oriented pack for Claude-style workflows, or the OpenAI-oriented pack for ChatGPT, Codex, or plugin-style workflows.
- 4
Run a small source-grounded task
Start with a simple research roadmap, authority-finding question, or citation check. Confirm the assistant can reach Descrybe Legal Engine and returns source material you can inspect.
- 5
Review before relying on the output
The workflows help structure research, but they do not provide legal advice or replace source review, jurisdiction checks, treatment review, or professional judgment.
Safety boundary
The workflow pack is designed to keep legal research structured and source-aware. It should help the assistant separate user-provided material, Descrybe results, and model reasoning.
- Identify jurisdiction assumptions before treating an answer as useful.
- Ask for Descrybe source results when the task depends on primary law.
- Flag missing facts, thin research, and unresolved citations.
- Avoid treating workflow output as legal advice.
- Review important sources directly before using the work in a filing, memo, client note, or public answer.