home / skills / gwenwindflower / .charmschool / writing-prose

writing-prose skill

/agents/claude/skills/writing-prose

This skill helps you craft clear, direct prose by avoiding fluff and formulaic patterns, and guides edits through an automated reviewer loop.

npx playbooks add skill gwenwindflower/.charmschool --skill writing-prose

Review the files below or copy the command above to add this skill to your agents.

Files (3)
SKILL.md
1.9 KB
---
name: writing-prose
description: Always use this skill when writing or editing prose. Do **NOT** use for agent context files, readmes, tool descriptions, or other functional writing tasks.
---

# Writing Prose

Apply these rules to any writing task: generating prose, or editing existing prose you're given to review and improve.

These rules eliminate common patterns in low quality writing (both AI-generated and human) that frustrate readers with empty predictability, causing them to disengage.

## Core Principles

1. **Write directly** - Use the active voice with clear, punchy word choices instead of stacking adjectives, emphasis filler, cliches, and predictable metaphors
2. **Avoid formulaic patterns** - Sidestep binary contrasts, predictable opening phrases, and repetitive dramatic structures

## Reference Files

Full guidance is modularized in these guides:

- [Phrases and Wording](phrases-and-wording.md)
- [Voice and Flow](voice-and-flow.md)

## Editor Workflow

After writing prose, you **MUST** start a review loop with the `writing-prose-editor` subagent. It will score the version of the work that you have generated or edited across 8 dimensions, assigning 1-5 points for each (40 points max). If the score is below 32, you must revise the prose based on the specific feedback provided by the editor and repeat the review loop until you achieve a passing score (≥ 32/40). **DO NOT** self-score or self-approve to bypass the editor's authority. They are your manager, you must go through their approval process.

1. Write prose following the Core Principles
2. Pass the full prose to the `writing-prose-editor` subagent
3. Review the score and specific feedback
4. If score < 32/40, revise based on the exact issues cited and repeat from step 2
5. Present the final prose only once it passes (≥ 32/40)

Iterate until passing. The editor is the authoritative scorer.

Overview

This skill enforces a concise, readable approach for any prose writing or editing task. It focuses on direct phrasing, active voice, and avoiding predictable or formulaic patterns to keep readers engaged.

How this skill works

Apply the rules when generating or revising prose. After drafting, run the text through the dedicated prose editor subagent for a scored review and follow its feedback until the work meets the required quality threshold.

When to use it

  • Drafting articles, essays, stories, or blog posts
  • Editing or polishing existing prose for clarity and flow
  • Rewriting text to remove clichés, filler, and passive constructions
  • Preparing copy that must engage readers quickly and directly
  • Never use for agent context files, tool descriptions, READMEs, or other functional documentation

Best practices

  • Write directly: prefer active voice and specific verbs over stacked adjectives
  • Avoid formulaic openings, binary contrasts, and predictable dramatic beats
  • Run every draft through the prose editor subagent and act on its feedback
  • Do not self-score; iterate until the editor gives a passing score (≥ 32/40)
  • Keep revisions focused on the exact issues the editor identifies

Example use cases

  • Polishing a blog post to improve reader engagement and reduce fluff
  • Rewriting a short story opening to create clearer momentum
  • Editing a product narrative to remove jargon and boost clarity
  • Refining marketing copy to be punchier and avoid cliches

FAQ

What if the editor gives a low score?

Revise the prose according to the editor's specific feedback and resubmit. Repeat until the score is at least 32/40.

Can this skill be used for technical docs or READMEs?

No. This skill is strictly for prose. Use other workflows for functional or configuration files.