Skip to content
SkillModel-led1,426 tokens in use

Voice-Preserving Rewriter

Rewrites text to remove AI-isms and returns the result in the chat, keeping the writer's voice, facts, intent, structure and protected material.

5,770 stars on the repository MIT Licenseconorbronsdon/avoid-ai-writing · skills/voice-preserving-rewriter
/plugin install avoid-ai-writing@conorbronsdon-skills
View source

Install command as written in the repository's README

Skill label

measured from its files

What installing Voice-Preserving Rewriter costs

47tokens idle
1,426tokens in use

Idle is what sits in context while the skill is installed. In use is its instructions, loaded when it runs. Counts are approximate, made offline with cl100k_base.

Model dependence Model-led

Script-ledMixedModel-led

Mostly instructions, so how well it works depends on the model that follows them.

This describes the skill's structure, how much of it is scripts. It does not say how well it works.

Touches

Nothing detected: no shell, network, file writes, git or credentials in its files.

Declared tags come from the skill's frontmatter; detected ones from its scripts, its code blocks and the commands it names in inline code. This is detected, not audited.

How it scores

Quality

How well the skill is written, counted from its own files: its frontmatter, its steps, examples, stated output and constraints, its size and whether every file it ships is one it names.How we score it →
Good
  • Frontmatter15/15
  • Structure23/25
  • Examples0/15
  • Output15/15
  • Size10/15
  • Resources0/15

Security

Starts at 100 and loses points for each thing in the skill's files that a reader should look at first.How we score it →
Excellent

No findings.

Repository health

Pre-computed score based on development activity, maintenance, community, maturity, and trend momentum.How we score it →
Excellent
  • Development activity96
  • Maintenance100
  • Community60
  • Maturity32
  • Momentum40

Voice-Preserving Rewriter handles requests to humanize, clean up or rewrite text when the result should come back as text and no file is changed. It follows the editing contract of the main Avoid AI Writing rulebook and edits only justified findings within the authorized scope.

Passages that already sound human stay as written, rough edges that belong to the writer’s style are kept, and quotes, code, tables and attributed text are not rewritten. The reply carries one final rewrite and a short verification note.

What You Get

  • One final rewrite of the text, with no superseded drafts
  • A concise change summary when it helps
  • A verification note covering passes used, checks run, residual findings and stop reason
  • Protected material, such as quotes, code, tables and URLs, left as written
  • A guard that keeps identity and representation details in prompts that describe people

Common Use Cases

  • Humanizing a LinkedIn post or newsletter draft
  • Removing AI-isms from an email while keeping the sender’s tone
  • Applying a casual or blunt voice to a rough draft
  • Fixing a rewrite that failed a preservation check, with one repair pass
  • Cleaning text that includes code or quotes that must not change

What Makes It Unique

It is built around not damaging the writer. It treats detector findings as evidence and not as orders, keeps source quirks that form the writer’s fingerprint, avoids swapping one cliche for another and forbids adding invented facts, experiences or confidence. It also caps editing at two passes and shares that budget with the verifier’s repair.

Under The Hood

The skill is instructions only, with no scripts. It loads one instruction file and follows the rulebook and routing files in the sibling skills. It returns text in the reply, does not write files, makes no network calls and runs no commands itself; the verifier it hands off to may run a Node script on the before and after text.

Join founders building with open source

Opinionated takes, migration guides, cost-saving tips, and insights from the open source ecosystem.

Subscribe on SubstackJoin 750+ subscribers