Skip to main content
skill CX Codex

llm-writing · haowjy, claude — Claude Code skill: what it does, how to install

llm-writing — a skill for Claude Code enforces writing guardrails on repository text. Strips default assistant phrasing from commits, PR descriptions, documentation, and prompt files.

Apache-2.0 815 installs ★ 601 updated 2026-11-16

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add haowjy/creative-writing-skills --skill llm-writing -a claude-code
Open on GitHub

What llm-writing does

This skill removes generic filler and synthetic assistant phrasing from durable repository files. It mandates short words, keeps exact project naming for symbols or flags, and cuts filler conclusions such as repetitive summaries.

Drafts are edited directly in the target file or a scratch pad before being committed. The instructions block conversational bleed, ensuring session IDs, chat-specific metaphors, and temporary working paths never reach PRs or permanent docs. For expanding pieces, it references companion skills intent-modeling and information-hierarchy.

What it helps with

  • strip repetitive model phrasing from commit messages and pull requests
  • remove ephemeral chat context from docs and AGENTS.md files
  • replace coined jargon with exact codebase terms
  • cut unevidenced assertions and empty summary paragraphs

Example prompts for the agent

  • Draft a pull request description following llm-writing rules
  • Edit the project README.md to remove default model filler
  • Write a commit message for the schema update without conversational bleed

How to install llm-writing

1Skills CLI (Vercel)

npx skills add haowjy/creative-writing-skills --skill llm-writing -a claude-code

2Manually: put the files in place

~/.claude/skills/llm-writing/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/llm-writing/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about llm-writing

How to install llm-writing in Claude Code?

Run npx skills add haowjy/creative-writing-skills --skill llm-writing -a claude-code. The file will be placed in ~/.claude/skills/llm-writing/SKILL.md.

What does llm-writing do?

It applies strict editorial guardrails to durable texts. The skill eliminates artificial transitions, unverified claims, and dialogue noise from commits, documentation, and prompts.

What does llm-writing require to run?

No external APIs or tokens are needed. For longer structures, it links with intent-modeling and information-hierarchy skills from the same repository.

Source text

SKILL.md · Apache-2.0
---
name: llm-writing
description: 'Load before you write a commit message, PR body, changelog entry, issue, report to the user, doc, README, AGENTS.md or .context/ page, KB page, or prompt: choose plain words, write for the reader''s context, and strip default LLM phrasing.'
metadata:
  type: guardrail
---

# LLM Writing

Load `/creative-writing-skills:intent-modeling` if it isn't already loaded, and `/creative-writing-skills:information-hierarchy` when the piece grows beyond a short answer.

## Write for the Reader

Text that outlives the conversation (files, prompts, docs, commit messages, PRs) is read by someone without it: a cold agent, a teammate, or the person who asked, after they've forgotten it. Before writing, decide who reads it, what they know when they start, and what they came for, then lead with what serves them and cut what only serves you. Break the piece into beats, each one move in the reader's journey. Check the source material it rests on.

Draft the whole piece where you can revise it before it reaches its reader: the file itself, or a scratch file. Text sent straight into a chat, a PR or a commit gets no revision. Revise from the whole piece down to the words: is each part correct, does it follow from what came before, and does the reader need it? After a local change, reread the beat around it.

## Choose Words

- Use the short, everyday word: "use", not "utilize"; "help", not "facilitate". A long word has to buy its length with precision.
- The project's own names are the word list. Write the real name as the project spells it, whether a symbol, a file, a flag, a character, or an invented term, not a synonym or a description of it, and call each thing by one name everywhere.
- Don't invent jargon. Use the words the reader would say out loud ("move", "delete", "a budget that only decreases"), not a coined metaphor ("evacuate", "ratchet"). A named pattern is fine when the text says what it means the first time.
- Be specific over sterile: not "schema changes can cause issues" but "a column rename fails the build".
- Vary the rhythm. Short sentences land a point; longer ones carry a fact with its condition or consequence. Split the sentence that carries two thoughts; keep the long one that carries one.
- Leave no sentence open to two readings: "only" and "not" next to the word they change, every "it" and "this" pointing at one noun, no noun strings. The cases and a before/after are in `resources/word-choice.md`.

## What to Delete

- Writing to fill a section because it exists. Delete it or merge its content where it belongs.
- Labeling concepts without explaining how they work. Explain the mechanism or cut the label.
- Stating conclusions without evidence. Show the evidence or drop the claim.
- Hiding uncertainty behind confident language. Say what you don't know.
- Softening every claim with qualifiers ("it's worth noting," "it's important to consider"). Say it or don't.
- Repeating what you already said in different words, or summarizing the body as a conclusion ("In summary," "Overall").
- Connecting ideas with transition words instead of meaning ("Moreover," "Furthermore," "Additionally"). If the relationship isn't clear without the word, restructure.
- Pairing clauses where one half already carries the meaning ("It's not X, it's Y"). Keep the half that carries it.
- Conversational bleed: writing as if replying inside the conversation that produced the text. Terms coined there, labels and IDs minted during the work (a ruling number, a finding ID, a lane name), the incident behind a rule, links into a work directory or anything else that won't last, "the old version," "as discussed," answers to questions the reader never asked. The reader wasn't there. Write for the reader, with reasons they can check without the conversation.

To commission a cold read of durable pages, and act on its findings, open `resources/cold-read.md`.

© haowjy · Apache-2.0 · haowjy/creative-writing-skills

Similar skills

All skills for Writing and docs →

llm-writing by other authors comes first.

llm-writing haowjy The llm-writing skill guides writing and revising text for people, from documentation to PRs. Use it to shape a draft around its reader and cut lines that add no useful information. CX Codex +1 815installs engineering-change-descriptions agentara engineering-change-descriptions is a Claude Code skill for drafting and rewriting PR descriptions, commit messages, and change summaries, including what changed, why, and how it was checked. CX Codex +1 124installs writing-commit-messages spencerpauly The writing-commit-messages skill gives the agent a fixed format for commit messages: type, scope, body, footer. Useful when the model writes the commits, not you. CX Codex +1 499installs llm-gate Rohit Ghumare llm-gate configures prompt and agent hooks in Claude Code to validate commit messages, stop dangerous terminal commands, and catch secret leaks before execution. CX Codex +1 ★ 2,966repo stars writing-user-docs callstackincubator writing-user-docs — a skill that sets voice, scope, and structure for user-facing docs. Use it when an agent writes or reviews anything a person reads to learn a tool. CX Codex +1 1,500installs writing-clearly-and-concisely davila7 writing-clearly-and-concisely — a Claude skill that pushes the agent toward Strunk's rules for clear prose and away from the stock phrasing language models fall into. CX Codex +1 206installs writing-simplified-technical-english PostHog writing-simplified-technical-english rewrites English prose to the ASD-STE100 controlled-language standard: one meaning per word, active voice, short sentences. Built for reports, PR descriptions and prompts. CX Codex +1 138installs

From the same set32

The whole set →
writing-principles The writing-principles Claude Code skill helps draft and assess fiction. It loads for prose writing, critique, or diagnosing why a passage feels flat. CX Codex +1 1,200installs creative-writing-craft creative-writing-craft guides fiction writing and line editing across production passes. It directs drafting, scene pacing, sensory grounding, POV, and style analysis. CX Codex +1 931installs llm-writing The llm-writing skill guides writing and revising text for people, from documentation to PRs. Use it to shape a draft around its reader and cut lines that add no useful information. CX Codex +1 815installs writing-staffing writing-staffing — a skill for Claude Code assigns story work to agents. It maps out their roles, the files and extra skills they need, and which models to use. CX Codex +1 787installs story-review story-review evaluates existing prose across five editorial levels, from plot structure down to copyediting. Useful when a draft needs diagnostic critique instead of automated rewrites. CX Codex +1 773installs story-planning story-planning — a Claude Code skill for creative direction, brainstorming, and story architecture before drafting prose. Activates when shaping plot beats or structuring scenes. CX Codex +1 763installs reader-sim reader-sim reads a draft through the eyes of a specified first-time reader and reports the experience as it unfolds. Use it when you want a reader’s response rather than a craft critique. CX Codex +1 634installs character-sim character-sim speaks as a specific fictional character during story moments, using their vocabulary, current emotional state, and knowledge limits. CX Codex +1 612installs

Project sponsors

Companies that keep this analytics open