agentleFS
Sign inStart free
DROOdotFOODROOdotFOO/agent-skills/skills/writing-voice/SKILL.md

writing-voice

Skill1 starsChanged 2 months agoMITView on GitHub

Writing voice calibration from studied authors. Accepts combinatorial voice arguments to blend multiple writer influences. TRIGGER when: user invokes "/writing-voice" or asks to "write like", "in the style of", "voice of", "blend voices", or references a specific writer influence for their prose. Also when user asks to review writing for voice/tone/style quality. DO NOT TRIGGER when: user asks to humanize existing text (use humanize skill), write code, generate commit messages, or produce technical documentation where voice is irrelevant.

We split this skill into the rules it gives an agent

8 rules. 0 are common, 7 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 9 edited versions we read keep these.

  • 7

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read the draft and flag where it diverges from the target voice.

    line 59
  2. Distinctive:

    Read several thousand words of the writer's prose across different topics

    line 63
  3. Very specific:

    Create voices/<name>.md following the template in voice-template.md

    line 64
  4. Distinctive:

    Add 3-5 real passages as calibration anchors

    line 67
  5. Distinctive:

    When rewriting with a voice profile loaded:

    line 71
  6. Distinctive:

    Read the full draft first.

    line 73
  7. Distinctive:

    Don't touch paragraphs that already have life.

    line 74
  8. Distinctive:

    After rewriting, read the full piece end-to-end.

    line 77

“Common” is measured: kept by at least 6 in 10 of the 9 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

writing-voice

Calibrate your writing voice against studied authors. Each voice file captures what a specific writer does -- sentence rhythms, structural habits, rhetorical moves, what they avoid -- so you can apply those patterns to your own prose without losing your identity.

How it works

Voice files in voices/ are reference profiles, not imitation targets. The goal is never to sound like Chapman or anyone else. It's to borrow specific moves -- a digression habit, a way of sitting with uncertainty, a paragraph rhythm -- and fold them into writing that's still yours.

Usage

Single voice

/voice chapman

Loads the Chapman voice profile. Apply his patterns where they improve the draft.

Blended voices

/voice pg-startup+chapman

Loads both profiles. Where they agree (e.g., both favor short pivot sentences after longer ones), lean hard into that. Where they conflict (PG opens with a thesis, Chapman opens mid-thought), pick per-paragraph based on what the material needs.

With intensity

/voice pg-startup:heavy chapman:light

Weight the influence. Heavy means structural and rhythmic borrowing. Light means just a few moves (e.g., borrow Chapman's habit of sitting with an uncertainty but not his digressive pacing).

Review mode

/voice review chapman

Read the draft and flag where it diverges from the target voice. Don't rewrite -- annotate. "This paragraph is in briefing mode. Chapman would sit with this idea longer." "This aside is good -- Chapman-like productive digression."

Adding a new voice

  1. Read several thousand words of the writer's prose across different topics
  2. Create voices/<name>.md following the template in voice-template.md
  3. Focus on moves, not content -- what they do structurally, not what they believe
  4. Include both what to borrow and what to avoid (every writer has tics)
  5. Add 3-5 real passages as calibration anchors

Applying voices to a draft

When rewriting with a voice profile loaded:

  1. Read the full draft first. Identify which paragraphs are working and which feel flat, mechanical, or rushed.
  2. Don't touch paragraphs that already have life. Voice calibration is for the dead spots.
  3. Apply moves from the loaded profile(s) -- but only where they serve the material. A Chapman-style digression in a status table is wrong. A Chapman-style moment of uncertainty in a risk section is right.
  4. Preserve the author's technical vocabulary and domain specificity. Voice is about rhythm, structure, and emotional register -- not word choice in technical passages.
  5. After rewriting, read the full piece end-to-end. Flag any sections where the voice shift is jarring (the "stitching" problem). Smooth transitions between the author's natural voice and the borrowed moves.

Available voices

VoiceFileBest for
Chapmanvoices/chapman.mdEssays, blog posts, anything that benefits from thinking-out-loud and productive digression
pg-earlyvoices/pg-early.mdCombative essays defending an unfashionable position; arguments that earn an abstract claim through a physical scene
pg-startupvoices/pg-startup.mdInstructional pieces, how-to writing, founder advice, any prose that lands on a portable imperative
pg-latevoices/pg-late.mdPhilosophical / cultural essays, refinement-driven arguments, pieces that show the writer changing their mind

What You Get

  • A calibrated writing voice applied to your draft, borrowing specific structural and rhythmic moves from loaded voice profiles
  • In review mode, annotated feedback showing where your prose diverges from the target voice without rewriting it
  • Blended voice configurations that weight multiple influences (heavy/light) per-paragraph based on what the material needs

Reading guide

TaskRead
Voice profile templatevoice-template.md
Chapman voice profilevoices/chapman.md
Paul Graham early (2001-2004)voices/pg-early.md
Paul Graham startup-era (2005-2015)voices/pg-startup.md
Paul Graham late (2016-present)voices/pg-late.md

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free