mcpField
H
human-writing

human-writing

github.com/timlai666Creative & media091 installsUpdated 2d ago
Repository

About human-writing

Human Writing is a skill for drafting or revising prose intended for people to read. It guides writers through choosing a mode, applying style principles, removing common AI writing patterns, and checking the result.

Key features

  1. 01Select a writing mode and audience
  2. 02Apply prose-writing principles
  3. 03Revise and generate prose
  4. 04Remove common AI writing patterns
  5. 05Check drafts against a checklist

Use cases

  1. 01Draft a project introduction
  2. 02Revise prose in a technical document
  3. 03Edit a README or repository description

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/timlai666/skills --skill human-writing

After installing, type/human-writingin Claude Code to confirm it loaded.

Bundled files

References and scripts installed along with SKILL.md.

agents/
openai.yaml
references/
ai-writing-patterns.md
final-checklist.md
human-writing-examples.md
human-writing-principles.md
mode-selection-and-language-rules.md

Details

From the project README

Human Writing

Overview

Use this for prose that a person will read, whether you are writing from scratch or revising a draft. The goal is prose the reader can picture and act on, with no AI tells left in.

Do not apply it to code, JSON, SQL, shell commands, or other structured output. When another skill owns the format or subject matter, use that skill to plan the content and use this one for the prose.

Project introductions, README files, and repository descriptions also use this skill. In mixed technical documents, apply the writing rules to prose while preserving the correctness of commands, code examples, and API identifiers.

This file gives the sequence. The rules are in the linked references. Read the reference named at each step instead of relying on a remembered summary.

Workflow

  1. Choose the mode. Read references/mode-selection-and-language-rules.md. Decide between rewrite and generate, identify the reader and purpose, match the language, and choose grounded or voiced.

  2. Learn the principles and examples. Read references/human-writing-principles.md, then read references/human-writing-examples.md in full before drafting. Study how the examples turn the rules into natural sentences, including rewrites that still leave too much for the reader to infer.

  3. Write the draft. Apply the principles and what the examples demonstrate about concrete detail, reader context, structure, vocabulary, rhythm, tone, and punctuation.

  4. Remove AI patterns. Read references/ai-writing-patterns.md. Remove every listed pattern without changing the meaning, facts, or intent.

  5. Check the finished text. Read references/final-checklist.md. Test every item against the actual text, fix failures, and check it again.

  6. Report ongoing task progress. At every turn end for any ongoing task, append a compact progress bar to the conversational reply, not the deliverable. Anchor it to the original goal: verified completed/total agreed items, current work, and what remains before completion. Show pending acceptance separately. Keep the denominator stable unless scope is explicitly changed; when the total is unknown, show milestones without a fabricated percentage or finish date. 100% requires all agreed completion and delivery conditions to be met.

Comments

Markdown supported