CommunityWriting & Editinggithub.com

jksim/agent-skills-jksim

Writes and edits text in simplified technical English, based on ASD-STE100. Use when writing docs, specs, plans, READMEs, comments or commit messages, or when the user asks for plain technical text.

What is agent-skills-jksim?

agent-skills-jksim is a Claude Code agent skill that writes and edits text in simplified technical English, based on ASD-STE100. Use when writing docs, specs, plans, READMEs, comments or commit messages, or when the user asks for plain technical text.

Works with✓Claude Code~Codex CLI~Cursor
npx skills add https://github.com/jksim/agent-skills-jksim/tree/HEAD/skills/simplified-english

Installed? Explore more Writing & Editing skills: steipete/notion, langchain-ai/langchain, bytedance/podcast-generation · View all 6 →

Ask in your favorite AI

Open a new chat with this agent skill pre-loaded.

Documentation

Simplified English

Overview

Rule 1 of Jarrett's rules: chat output, generated docs, code comments and commit messages use simplified technical English. The style is based on the principles of ASD-STE100. Readers find facts faster. Readers whose first language is not English understand more.

This skill is our own summary. It is not the ASD-STE100 standard and does not include its dictionary.

When to Use

  • You write or edit a spec, plan, ADR, README, vocabulary entry or other doc.
  • You write a code comment or a commit message.
  • You review text for rule 1.

Do not apply it to code identifiers, quoted output (errors, logs, command output), skill files or text the user wrote.

Process

Write the text. Then check it against each rule.

#RuleNot thisThis
1One topic per sentence.The hook reads the file and, if it is missing, it prints a warning, which the user sees.The hook reads the file. If the file is missing, the hook prints a warning. The user sees the warning.
2Instructions: 20 words or fewer. Descriptions: 25 words or fewer.Make sure that before you go ahead and push the branch you have run the full test suite and checked the output.Run ./scripts/test.sh. Then push the branch.
3Use the active voice.The rules are loaded by the hook.The hook loads the rules.
4Instructions use the imperative.The test should be run before the commit.Run the test before the commit.
5One word has one meaning. Use the terms in .claude-work/docs/VOCABULARY.md.Warn, then later "flag", "alert" or "notify" for the same action.Warn, every time.
6Use simple tenses: present, past, future.The hook would have been checking the input.The hook checks the input.
7Use a single verb, not a phrasal verb or idiom.Kick off the build. Figure out the cause.Start the build. Find the cause.
8Keep "the" and "a".Run script, check output.Run the script. Check the output.
9One instruction per step.1. Open the file and change the version.1. Open the file. 2. Change the version.
10Paragraphs: 6 sentences or fewer.A block of ten sentences.Two short paragraphs, or a list.

More rules:

  • Write a warning or caution before the step it applies to. Start with the danger: "Warning: this deletes the branch."
  • Use numbers, not words, for values: "72 characters", not "seventy-two characters".
  • Use a list or a table when you give three or more items.
  • Do not use synonyms to vary the text. Repeat the same term.
  • Do not use "it" or "this" when the noun is not clear. Name the noun again.
  • When a term is not in the vocabulary and the meaning is not clear, ask the user. Then add the term to the vocabulary.

Common Rationalizations

RationalizationReality
"Short sentences sound robotic."Short sentences are fast to read and hard to misread. The reader's speed matters more than style.
"A synonym makes the text less repetitive."A synonym makes the reader ask if it is a different thing.
"This comment is too small for the rules."Comments and commit messages are in scope. Small text is easy to fix.
"The passive voice is more formal."The passive voice hides who does the action. Name the actor.

Red Flags

  • Sentences with "and", "which" or "so that" that join two topics.
  • "Should", "could" or "might" in an instruction.
  • The same thing has two names in one doc.
  • A step with two actions.
  • A warning after the step it applies to.

Verification

  • Each sentence has one topic.
  • No instruction is longer than 20 words. No description is longer than 25 words.
  • All instructions use the imperative and the active voice.
  • All terms match the vocabulary.
  • Each list step has one instruction.
  • No paragraph is longer than 6 sentences.

Individual skills in this repo

This repo contains 2 individual skills — each has its own dedicated page.

Related Skills