# Polish Drafts While Preserving Your Voice

Automated drafts can sound repetitive or overly polished and may lose the writer's original voice. This skill applies minimal editorial fixes while preserving meaning.

## Install

```bash
npx skillstore add glenskii/human-writer
```

## Metadata

- Status: approved
- Slug: glenskii-human-writer
- Version: 1.2.0
- Author version: 1.2.0
- Skillstore revision: r2
- Version status: valid
- Tree hash: e2c4c1c47ab09605a7336b0474bc474d5c710e5a408c44853259ac0c9a18918a
- Author: glenskii
- GitHub username: glenskii
- License: MIT
- Repository: https://github.com/glenskii/glenski-plugins/tree/60a0fdb8645aff907f29880bc70b1ee62ec7309c/plugins/glenski-quality-suite/skills/human-writer
- Ref: f4838806900353c20a6899b1cbb5f5bc3a23357a
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands
- Quality score: 90
- Quality tier: featured
- Public page: https://skillstore.pages.dev/skills/glenskii-human-writer
- Manifest: https://skillstore.pages.dev/api/skills/glenskii-human-writer/manifest

## Capabilities

- Edits existing drafts without adding new claims, examples, opinions, or arguments.
- Detects listed machine-written patterns without assigning an authorship score.
- Preserves visible voice traits, formatting, factual boundaries, and approved terminology.
- Applies different editing depth for social posts, documentation, correspondence, and public copy.
- Flags uncertain factual or stylistic choices instead of silently changing them.
- Checks public-facing text for prohibited terms and long dashes with a bundled local script.

## Use Cases

- Refine a professional email: Remove robotic phrasing and formatting artifacts while keeping the sender's directness, facts, and requested action unchanged.
- Review technical documentation: Apply the full checklist while protecting precise claims, terminology, headings, and the document's established register.
- Audit a social draft: Identify visible writing patterns without rewriting the post or making unsupported claims about authorship.

## Prompt Templates

### Quick draft cleanup

```
Edit this draft with the lightest possible touch. Remove clear machine-written patterns, preserve my meaning and formatting, and return only the corrected text.
```

### Detect patterns without editing

```
Audit this draft without rewriting it. Quote each detected pattern, name the checklist category, and suggest a brief fix without scoring authorship.
```

### Preserve a defined voice

```
Edit this draft using the supplied voice profile. Preserve its tone, rhythm, terminology, facts, and deliberate rough edges. Flag any uncertain change.
```

### Prepare public documentation

```
Apply the full documentation and public-copy checks to this file. Preserve technical meaning, run the bundled editorial checker, and summarize substantial change categories.
```

## Limitations

- It does not generate new copy from scratch or expand an argument.
- It does not verify facts beyond flagging an obvious possible error.
- It cannot determine whether a person or model authored a text.
- Its fixed pattern list may flag intentional style choices that require writer review.

## Best Practices

- Provide the complete draft and identify its destination, such as email, documentation, or social media.
- Supply a short voice profile when punctuation, terminology, or brand language must remain consistent.
- Review flagged factual and stylistic choices because the skill intentionally leaves uncertain changes unresolved.

## Anti Patterns

- Do not ask the skill to create a new article, argument, example, or factual claim.
- Do not use detection results as proof that a model authored third-party text.
- Do not request aggressive rewriting when preserving the writer's meaning and voice is required.

## Security Audit

- Audited at: 2026-08-09T08:15:04.242\+00:00
- Summary: All 15 static findings are false positives caused by editorial prose, Markdown code spans, or a clear Unicode literal. The bundled script performs local text checks without dynamic commands, network access, or obfuscated behavior.

## Stats

- Views: 4
- Downloads: 26
- Favorites: 1
- Popularity score: 0