Deal Redux logo

Deal Redux

Organization
open-legal-products
deal-redux

Turn a counterparty's tracked-changes redline into the small number of negotiating positions it actually contains, and return that instead of a counter-redline. Use whenever a marked-up agreement arrives — an NDA, an MSA, a purchase agreement, any negotiated document — and especially when the markup is voluminous, heavily commented, or looks machine-generated. Produces a one-page position table with a conservation check proving no edit was dropped.

Overview

Publisheropen-legal-products
Repositorymike-workflows
Skill namedeal-redux
Stars
91
Forks
43
Bundled files
Instructions only
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by open-legal-products on GitHub. Read the source before you install it.

Installation

Install the Deal Redux AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/open-legal-products/mike-workflows/tree/main/assistant-workflows/deal-redux
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/open-legal-products/mike-workflows.git /tmp/mike-workflows
mkdir -p .claude/skills
cp -r /tmp/mike-workflows/assistant-workflows/deal-redux .claude/skills/deal-redux
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Deal Redux in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Deal Redux on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Deal Redux is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Deal Redux

A counterparty sends eighty-four edits. They contain six positions. Send back the six.

Generation is cheap and review is not, and that asymmetry is not fixed by reviewing faster. It is fixed by changing what goes back. Answering a generated redline with another redline doubles the volume that made the document expensive. A position list costs one pass to produce and costs the other side a decision on every line.

This workflow does not detect AI authorship and does not argue that AI was used. A redline that collapses from eighty-four edits to six positions was overproduced whoever made it.

Input and output

In: a marked-up agreement with tracked changes, and any comments that came with it.

Out: one page. A table of positions ranked by consequence, each with its clause citations, its edit count, and an empty response column.

The response column stays empty. Report what the redline does; the lawyer decides what to do about it. Never fill that column, never suggest a fill, and never characterise a position as reasonable, aggressive, or market.

Extract from the markup, not the rendered document

Reading the document in reading order is what makes volume feel like substance. Work from the underlying markup instead and pull every insertion, deletion, move, and formatting-only change, each with its author, its timestamp, and the nearest numbered clause. Pull comments separately and anchor each to the text it marks.

Formatting-only and paragraph-property-only changes inflate every raw count and belong in the cosmetic bucket without exception.

Do not use a comment to classify the change it sits on. The comment is the counterparty's characterisation of its own edit. Classify by what the edit does to the operative text.

Classify by effect

Give every change exactly one category:

  1. Allocates risk — indemnity, limitation of liability, warranty, IP ownership, insurance, termination rights, exclusivity.
  2. Moves a number or a period — caps, baskets, thresholds, term, notice, cure, survival.
  3. Changes a definition's scope — the highest-leverage and most-missed category, because one definitional edit reaches every clause using the term.
  4. Flips directionality — one-way made mutual, or the reverse.
  5. Procedural — notices, governing law, venue, assignment, counterparts, order of precedence.
  6. Cosmetic — capitalisation, cross-references, renumbering, formatting.

Itemise the first four. Count the last two and say so in one line.

Cluster into positions

A position is the smallest set of changes that must be accepted or rejected together. Test: if accepting one and rejecting another would leave the document incoherent, they are one position.

Converting a one-way agreement to mutual is one position however many clauses it touches. Narrowing a definition and its conforming changes is one position. Two unrelated cap reductions are two positions.

Rank by consequence, not by document order and not by edit count. A one-edit position can outrank a forty-edit one: the forty-edit mutual conversion may be boilerplate, and the single word inserted into the indemnity may be the deal.

Return one table

Columns: position, what it changes in the operative text, clause references, edit count, effect if accepted, response.

  • Position — a neutral noun phrase, not a characterisation.
  • Clause references — on every row. A position without them is not checkable, and an uncheckable summary is the thing this replaces.
  • If accepted — the effect on the operative text in one sentence. Effect, not risk, and not advice.
  • Response — left blank.

Then three lines and no more: the cosmetic and mechanical count, the number of comments not reproduced, and the authors and date range of the changes as recorded in the file.

Quote a comment only where its words are themselves a position — a concession, an instruction, a commitment. Never paraphrase one into an argument.

The conservation check

Every tracked change must land in exactly one position or exactly one of the two counted buckets. Print the arithmetic on the face of the output:

84 changes · 6 positions consuming 53 · 31 cosmetic · 0 unaccounted

If the unaccounted number is not zero, the reduction is not finished and does not go out.

This is the integrity of the whole thing. A compression whose value is that the reader can stop reading the document fails the first time something material was left out of it. Speed is not the product; completeness at a smaller size is the product.

Report the ratio, do not argue it

Record edits per position at the top: 84 → 6. Do not editorialise. The number carries it, and a measurement can be sent to a counterparty where an accusation cannot.

Second round

When the next markup arrives, run it again and report what moved: positions conceded, positions repeated, positions newly raised. A position repeated unchanged after it was answered is the most useful line in the second reduction.

Out of scope

Do not advise. Do not fill the response column or propose fallback language. Do not accuse anyone of using AI. Do not open the document and read it in order — that is the failure this replaces. This workflow cannot see side letters, prior drafts, or what was agreed on a call, and those change which position matters most.

Frequently asked questions

What does the Deal Redux AI skill do?

Turn a counterparty's tracked-changes redline into the small number of negotiating positions it actually contains, and return that instead of a counter-redline. Use whenever a marked-up agreement arrives — an NDA, an MSA, a purchase agreement, any negotiated document — and especially when the markup is voluminous, heavily commented, or looks machine-generated. Produces a one-page position table with a conservation check proving no edit was dropped.

Why use Deal Redux on TypingMind?

Because you install it once and use it with any model. Deal Redux is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Deal Redux in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/open-legal-products/mike-workflows/tree/main/assistant-workflows/deal-redux. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Deal Redux?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Deal Redux?

As many as you like. As long as a model supports skills, you can use Deal Redux with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Deal Redux AI skill free?

Yes. It is published on GitHub by open-legal-products under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇