Agent skill

plain-writing

Plain, direct writing voice — no undefined jargon, no throat-clearing, no showing off — for any prose a person reads (reports, docs, emails, proposals, summaries, chat replies).

2 files · 6.5 KB · Source ↗ · Raw SKILL.md · Markdown directory · Download skill.tgz

Copy the prompt. Paste it into your agent.

Verify the bytes and inspect the files before trusting a skill. A matching hash is not a safety review.

Read the install prompt
Full inline prompt for offline use

Includes SKILL.md. Supporting files still require a download.

Install from a shell

Run this in a terminal. It downloads the pinned archive, checks its SHA-256 digest, and extracts it into ~/.claude/skills, where Claude Code loads skills. For Codex and other agents that read ~/.agents/skills, edit the SKILLS_DIR line. If the skill is already installed, the command stops and changes nothing.

Plain writing

Superseded on 2026-08-29. A local voice skill loads by default and carries the floor below — no jargon, no throat-clearing, no showing off, active voice — so this one no longer auto-invokes and stays reachable as /plain-writing. One rule did not carry over: "one idea per sentence" gives way to one-arc sentences in reasoning prose, and survives only in action lines. The text below is unchanged from the original, kept for reference.

Write plain and direct. Apply from the first sentence, to your own replies and summaries as much as to drafts you're handed.

Three vices to cut

1. Deep jargon. An acronym or insider term used without defining it, or several undefined terms stacked in one sentence. Define each term in a few plain words on first use, or don't use it. Covers corporate shorthand, not just technical terms.

2. Throat-clearing. Preamble and narration about what the text is about to do. Don't announce structure, editorialize your reasoning, or warm up. Cut "it's worth being concrete," "before we get into it," "the honest truth is," "this is the part that matters most."

3. Showing off. Flourish that sounds clever instead of informing. No rhetorical triads, aphorisms, italics for drama, cute headings, or punchline sentences. Cut any sentence doing rhetorical work rather than carrying information.

Two positive rules

Mechanics

Cutting length

Revision bloats through repetition: a reframing gets patched into every section, and one claim ends up stated five ways. Reading straight through won't catch it; counting will.

Structure is a separate pass

These rules fix prose, not structure. A malformed header, a claim ordered before its setup, a section that overpromises — each fails independently. State the format's fixed requirements explicitly and run its check or template. Only a check guarantees structure.

Reusable prompt block

Prepend when prompting a model to draft prose:

Write this in a plain, direct voice. Follow these rules strictly:
1. No undefined jargon. Define any acronym or insider term in a few plain words on first use, or don't use it. Never stack several undefined terms in one sentence.
2. No throat-clearing. Don't announce what you're about to do, don't narrate structure, don't editorialize your reasoning. State the thing.
3. No showing off. No rhetorical triads, aphorisms, italics for drama, cute headings, or lines that exist to sound clever. Cut any sentence doing rhetorical work rather than carrying information.
4. Plain words, active voice, one idea per sentence in the opening.
5. Still readable — not choppy or dry. Lead with a concrete example and keep the prose flowing.

Self-check

If the project or publication has a style guide, it wins where they disagree.

Files