A portable skill for AI coding assistants and agent frameworks (including Google Antigravity, Claude Code, Codex CLI, and others). It teaches the agent to audit, draft, and rewrite text into plain English, for legal writing and general prose alike.
It applies the core rules of plain-language legal drafting (inspired by Bryan Garner's Legal Writing in Plain English) to turn dense legalese and verbose AI slop into clear, readable prose.
Built for lawyers, paralegals, in-house counsel, and law students. It works on contracts, briefs, memos, opinions, client letters, demand letters, and policies. It also suits anyone cleaning up AI-generated text before it goes to a client or a court.
Note
Not only for lawyers: The skill protects legal terms of art and quoted language. Its core rules (sentence architecture, active voice, needless words) apply to any professional writing: business memos, documentation, and correspondence. The same rules cut the padding and repetition in AI-generated text ("AI slop").
- Legalese Buster: Replaces archaic terms (aforesaid, hereinafter, said, such) and redundant doublets/triplets (null and void, give, devise, and bequeath).
- Sentence & S-V-O Optimization: Reconstructs long, winding sentences to place subjects and verbs close together near the beginning, keeping the writer's own sentences to a readable 20-word average.
- Quotations Stay Verbatim: Never paraphrases quoted contract, statutory, or case language. It shortens a quotation only with a Bluebook ellipsis or bracketed alteration, and it edits the prose around the quotation instead.
- Professional Register: Replaces slang and colloquial idioms with precise words of the same meaning, while leaving figurative terms of art (on all fours, safe harbor) alone.
- Nominalization De-bloater: Converts buried verbs (like make a decision or conduct an analysis of) back into strong, active verbs (decide, analyze).
- Deep Issue Framing: Restructures complex legal issues into the clean "premise-premise-question" format (under 75 words) rather than convoluted "Whether" statements.
- Rule Tabulation: Formats dense multi-part legal rules, conditions, and lists into structured, parallel, and indented sub-items, and keeps short, simple series inline.
- Double-Entry Number Clean-up: Eliminates parenthetical repetition (e.g., three (3) days -> 3 days) and uses digits for every figure a reader may act on: dates, money, percentages, section numbers, and deadlines.
- Gender-Neutral Drafting: Replaces binary/clunky pronouns (he or she, his or her) with gender-neutral phrasing.
- Oxford Comma Enforcement: Uses the serial comma to prevent list ambiguity.
- Omitting Needless Words: Strips redundant modifiers (e.g., completely finish -> finish), empty "throat-clearing" introductory phrases, and roundabout sentences.
- ABA Publication Style Compliance: Standardizes restrictive/nonrestrictive clauses (that vs. which), quotation punctuation, date/name formatting, U.S. abbreviations, and "attorney fees" terminology.
- Court Rules First: Applies its typography defaults (serif body, ragged right, no all-caps) only to attributes that no court rule, agency rule, or house style prescribes.
- Structural Formatting: Redesigns long, dense walls of legal text using informative headings and lists.
This skill uses the portable SKILL.md (Agent Skills) format, natively supported by Antigravity CLI (Google's successor to Gemini CLI), Claude Code, and Codex CLI. See the framework-specific instructions below.
The fastest path. Works across 20+ agents through the skills.sh registry:
npx skills add fayerman-source/deslopAntigravity CLI replaced Gemini CLI on June 18, 2026 (announcement) and keeps native Agent Skills. It discovers a SKILL.md placed in its skills directories; clone the repository into one:
- Workspace-specific:
git clone https://github.com/fayerman-source/deslop.git .agents/skills/deslop
- Global (all workspaces):
git clone https://github.com/fayerman-source/deslop.git ~/.gemini/skills/deslop
The CLI discovers the skill automatically. If it doesn't appear, restart the CLI or reload skills. Antigravity's native install command may still change, so check the official docs for current syntax. The manual clone above works either way.
Gemini CLI (Code Assist Standard and Enterprise only): If you still run Gemini CLI, it installs the same skill with
gemini skills install https://github.com/fayerman-source/deslop.git(append--scope workspacefor project scope).
Claude Code supports this skill natively through its Agent Skills system. SKILL.md already carries the required name/description frontmatter. Clone the repository into a skills directory:
- Personal (all projects):
git clone https://github.com/fayerman-source/deslop.git ~/.claude/skills/deslop - Project-specific:
git clone https://github.com/fayerman-source/deslop.git .claude/skills/deslop
Claude Code auto-discovers the skill and applies it when relevant; you can also invoke it explicitly with /deslop.
Inside a session: Claude Code has no slash command to install a standalone skill from a URL. Either run the clone above, or ask Claude to do it for you ("clone deslop into
~/.claude/skills"). Restart Claude Code afterward so it picks up the new skill.
Alternative (without skills): Append the contents of SKILL.md to a CLAUDE.md file: your project-root CLAUDE.md for a single project, or ~/.claude/CLAUDE.md for all projects. Claude Code loads both automatically as instructions.
Codex CLI also supports native Agent Skills. It scans .agents/skills/ directories for a SKILL.md with name/description frontmatter (keep the frontmatter; Codex uses it for discovery). Clone into a skills directory:
- Project-scoped:
git clone https://github.com/fayerman-source/deslop.git .agents/skills/deslop
- User-scoped (all repositories):
git clone https://github.com/fayerman-source/deslop.git ~/.agents/skills/deslop
Invoke it with $deslop, or let Codex activate it when the task matches the skill description. (Restart Codex if a freshly cloned skill doesn't appear.)
Inside a session: use the built-in
$skill-installerskill: run$skill-installerand ask it to install deslop from the repository URL (it can pull skills from other repositories). Browse what's installed with/skills.
Alternative (always-on): copy the body of SKILL.md (without the YAML frontmatter) into AGENTS.md at your repo root, or ~/.codex/AGENTS.md for global guidance. Codex reads these as standing instructions.
Tools without a skills system can still use these guidelines; load SKILL.md as the tool's instructions or rules file:
- Aider:
aider --read SKILL.md(or addread: SKILL.mdto.aider.conf.yml). - Cursor: copy
SKILL.mdinto.cursor/rules/. - GitHub Copilot: paste the rules into
.github/copilot-instructions.md.
Once installed, your AI agent will automatically detect and apply the skill when you ask it to edit legal documents, or you can invoke it explicitly:
-
Explicit Prompting:
"Review this contract clause using the
deslopskill." "Draft a brief response to this motion following thedeslopguidelines." -
Slash Commands: If your agent supports slash commands in its TUI, you can trigger it directly:
/deslop Audit this NDA agreement: [paste agreement text] -
What you get back: an audit table with one row for each rule section (IβVI), so you can see what the agent checked, not only what it found. Sections with no violations say "none found." The Section III row lists the terms of art the agent kept verbatim, and the Section I row lists the quotations it kept. The rewrite follows the table.
In the event that the Party of the Second Part fails to make payment of any sum due hereunder subsequent to the receipt of written notice of default, the Party of the First Part may, at its option, declare this Agreement null and void and of no further force and effect, and may proceed to exercise any and all remedies available to it under the law.
If Jones fails to pay any amount due after receiving a written default notice, Smith may terminate this agreement and pursue any legal remedies.
Contributions are welcome: improve style guidance, expand rules, or add support for more agent frameworks. Open an issue or submit a pull request.
Distributed under the MIT License. See LICENSE.
