Installable skillMIT

unslop — Strip AI Writing Patterns via CLI

Post-process AI-generated text through the unslop CLI to strip AI writing patterns before publishing

BY FrancoStinoUPDATED Oct 28, 2026
INTENDED FOR✳Claude Code>_Codex◈OpenClaw↗Cursor✦GeminiInstall in a listed agent.

instructions.md

---
name: unslop-ec35dfc1b090
description: "Post-process AI-generated text through the unslop CLI to strip AI writing patterns before publishing"
category: writing
risk: safe
source: community
source_repo: MohamedAbdallah-14/unslop
source_type: community
date_added: "2026-04-25"
author: MohamedAbdallah-14
tags: [writing, content-quality, ai-writing, text-processing, cli, publishing]
tools: [claude-code, cursor, gemini-cli, codex-cli, antigravity]
license: "MIT"
license_source: "https://github.com/MohamedAbdallah-14/unslop/blob/main/LICENSE"
---

# unslop — Strip AI Writing Patterns via CLI

## Overview

unslop is a CLI tool that post-processes text to remove AI writing patterns programmatically. Unlike skills that ask the agent to avoid AI-isms, unslop runs as a deterministic pipeline step: pipe text in, get clean text out. Use it as a final pass before committing docs, publishing posts, or sending any AI-generated content to production.

The `--deterministic` flag makes output reproducible — same input always produces same output. The `--stdin` flag reads from stdin, enabling shell pipeline composition.

## When to Use This Skill

- When you have AI-generated text ready to publish and want a final cleanup pass
- When working in a shell pipeline where text quality needs to be enforced automatically
- When writing commit hooks or CI steps that validate content before it ships
- When you need reproducible text normalization across multiple runs

## Setup

Install once:

```bash
pipx install unslop
# or
uv tool install unslop
```

Verify:

```bash
unslop --version
```

## How It Works

### Step 1: Pipe Text Through unslop

Standard cleanup (may vary slightly between runs):

```bash
echo "This leverages cutting-edge AI to deliver robust solutions." | unslop --stdin
```

Deterministic cleanup (same input → same output every run):

```bash
echo "This leverages cutting-edge AI to deliver robust solutions." | unslop --stdin --deterministic
```

### Step 2: Use in Shell Pipelines

Pipe the output of any command through unslop:

```bash
cat draft.md | unslop --stdin --deterministic > clean.md
```

Or chain with other tools:

```bash
cat draft.md | unslop --stdin --deterministic | pbcopy   # macOS: copy clean text to clipboard
```

### Step 3: Integrate into Commit Hooks or CI

Add to a pre-commit hook or CI step to enforce quality gates on any generated content before it ships:

```bash
# In .git/hooks/pre-commit or a CI script
CONTENT=$(cat docs/changelog.md)
CLEANED=$(echo "$CONTENT" | unslop --stdin --deterministic)
if [ "$CONTENT" != "$CLEANED" ]; then
  echo "Changelog contains AI writing patterns. Run: cat docs/changelog.md | unslop --stdin --deterministic > docs/changelog.md"
  exit 1
fi
```

## Examples

### Example 1: Clean a Draft Document

```bash
cat blog-post-draft.md | unslop --stdin --deterministic > blog-post-final.md
```

### Example 2: Inline Cleanup During Writing

```bash
# Write content, pipe through unslop, write result back
cat README.md | unslop --stdin > README.clean.md && mv README.clean.md README.md
```

### Example 3: Validate Before Submitting a PR

```bash
# Check if any generated docs need cleanup
for f in docs/*.md; do
  ORIGINAL=$(cat "$f")
  CLEANED=$(echo "$ORIGINAL" | unslop --stdin --deterministic)
  [ "$ORIGINAL" != "$CLEANED" ] && echo "Needs cleanup: $f"
done
```

## Best Practices

- ✅ Use `--deterministic` in CI and automation to ensure reproducible output
- ✅ Run on the final draft, not intermediate iterations
- ✅ Combine with the `avoid-ai-writing` skill for both generation-time guidance and post-processing
- ❌ Don't run on code files — unslop targets prose, not source code
- ❌ Don't skip review after unslop: automated cleanup can occasionally change meaning; read the output

## Limitations

- Processes prose only — not code, JSON, or structured data
- Does not catch factual errors or substantive writing issues
- Some replacements may not fit every context; review the output before publishing
- Requires Python tooling such as `pipx` or `uv` for standalone CLI installation

## Security & Safety Notes

- unslop reads from stdin and writes to stdout — no file system side effects by default
- `--deterministic` mode is local and does not make LLM API calls
- Default LLM mode may use `ANTHROPIC_API_KEY` or the Claude CLI; use `--deterministic` for sensitive local files and CI gates
- Safe to run in CI pipelines and commit hooks when pinned to deterministic mode
COMPATIBILITY & REFERENCES

Platform guidance

Imported from the listed AgentSkill.sh category. Complete supplied package files and notices are retained. Required external tools and credentials are not installed. No bundled code was executed during import. Individual agent task behavior has not been tested.