Humanize markdown and text files — strip AI-isms while preserving code, URLs, headings, and technical content.
Project description
unslop
Your CLAUDE.md sounds like ChatGPT wrote it. Stop that.
A bundled unslop skill that rewrites memory files (CLAUDE.md, AGENTS.md, todos, preferences, docs) to remove AI slop while preserving every code block, URL, path, and heading exactly.
What It Do
Slash command: /unslop-file <filepath>
Drops sycophancy, stock vocab (delve, tapestry, testament, leverage, robust, holistic), hedging stacks, tricolon padding, em-dash pileups, and tidy five-paragraph essay shapes. Engineers burstiness. Restores voice. Keeps every fenced code block byte-identical.
Backup: FILE.original.md (the human-readable original — re-run if you edit it).
Two Modes
Deterministic (fast, no API)
cd unslop && python3 -m scripts --deterministic <absolute_filepath>
Pure regex. No API key. Strips canonical AI-isms and tightens tricolons. Best for batch / CI. ~50ms per file.
LLM (default, calls Claude)
cd unslop && python3 -m scripts <absolute_filepath>
Calls Anthropic API (if ANTHROPIC_API_KEY set) or claude --print CLI. Better quality: rewrites for burstiness, restructures performative paragraphs, matches voice when sample provided.
Benchmarks
Sample memory files, before/after AI-ism count and word count (deterministic mode):
| File | Original words | Humanized words | AI-isms removed |
|---|---|---|---|
claude-md-preferences.md |
706 | 624 | 18 |
claude-md-project.md |
1,512 | 1,310 | 24 |
mixed-with-code.md |
950 | 838 | 11 |
project-notes.md |
1,180 | 1,022 | 19 |
todo-list.md |
845 | 763 | 14 |
LLM mode produces stronger rewrites (typically 60–75% AI-ism removal and noticeable burstiness improvement) but takes ~5–15s per file and costs API tokens.
Security
unslop calls a subprocess (the claude CLI) when ANTHROPIC_API_KEY is unset. Snyk may flag this as High Risk because of the subprocess.run + file I/O combination. See SECURITY.md for the full rationale.
The deterministic mode does no subprocess and no network — use it if your security context requires that.
Install
Bundled with the unslop plugin. Skill path: unslop/.
Standalone: cd unslop && python3 -m scripts <filepath> (Python 3.10+).
Usage
# Deterministic, no API
python3 -m scripts --deterministic /Users/me/proj/CLAUDE.md
# LLM, default
ANTHROPIC_API_KEY=... python3 -m scripts /Users/me/proj/CLAUDE.md
# LLM via Claude CLI fallback
python3 -m scripts /Users/me/proj/CLAUDE.md
What Files Work
| Type | Action |
|---|---|
.md, .markdown, .txt, .rst |
Humanize (prose) |
.py, .js, .ts, .json, .yaml, etc. |
Skip (code/config) |
*.original.md |
Skip (already a backup) |
| Mixed prose + code | Humanize prose; leave fenced code untouched |
Sensitive paths (.env, *.pem, ~/.ssh/) |
Refuse |
How It Work
detect file type ──▶ write .original.md backup ──▶ humanize (deterministic or LLM)
│
▼
validate
(preserve code/URLs/paths + AI-ism residual)
│
┌──────┴──────┐
pass fail
│ │
write & done targeted fix
│
retry (max 2)
│
┌───────┴───────┐
pass fail
│ │
write & done restore original
What Is Preserved
- Fenced code blocks (every byte)
- Indented code blocks
- Inline backticks
- URLs and markdown links
- File paths and commands
- Technical terms, proper nouns, API names
- Dates, version numbers, numerics
- Headings (exact text)
- Tables (structure; cells humanized)
- YAML frontmatter
Why This Matter
Project memory files (CLAUDE.md, AGENTS.md) get loaded into every agent session. If they're in AI-template English, you pay tokens for slop and you teach the model to mirror that voice back at you. Humanizing the memory file is the cheapest possible voice intervention.
Part of the Unslop Plugin
| Skill | What it does |
|---|---|
unslop |
Active humanization mode for live responses |
unslop-commit |
Commit messages without AI tone |
unslop-review |
Direct, human-voice PR comments |
unslop |
This: rewrite memory files |
unslop-help |
Quick-reference card |
License
MIT — same as the parent unslop plugin.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file unslop-0.5.0.tar.gz.
File metadata
- Download URL: unslop-0.5.0.tar.gz
- Upload date:
- Size: 54.7 kB
- Tags: Source
- Uploaded using Trusted Publishing? Yes
- Uploaded via: twine/6.1.0 CPython/3.13.12
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
e326510fd74fb4e48694c7a688f9e8f652c8c3733b873316833104b879955bb1
|
|
| MD5 |
379f8f5eac051e74181fc4b910351e1a
|
|
| BLAKE2b-256 |
8eac19c977af125c53793e396778b9d4be69ce79cf8040b49054c48b66239dbf
|
Provenance
The following attestation bundles were made for unslop-0.5.0.tar.gz:
Publisher:
publish.yml on MohamedAbdallah-14/unslop
-
Statement:
-
Statement type:
https://in-toto.io/Statement/v1 -
Predicate type:
https://docs.pypi.org/attestations/publish/v1 -
Subject name:
unslop-0.5.0.tar.gz -
Subject digest:
e326510fd74fb4e48694c7a688f9e8f652c8c3733b873316833104b879955bb1 - Sigstore transparency entry: 1352751351
- Sigstore integration time:
-
Permalink:
MohamedAbdallah-14/unslop@444a3821649e42fd55240d58bdaba609894fa9e9 -
Branch / Tag:
refs/tags/unslop-v0.5.0 - Owner: https://github.com/MohamedAbdallah-14
-
Access:
public
-
Token Issuer:
https://token.actions.githubusercontent.com -
Runner Environment:
github-hosted -
Publication workflow:
publish.yml@444a3821649e42fd55240d58bdaba609894fa9e9 -
Trigger Event:
push
-
Statement type:
File details
Details for the file unslop-0.5.0-py3-none-any.whl.
File metadata
- Download URL: unslop-0.5.0-py3-none-any.whl
- Upload date:
- Size: 58.6 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? Yes
- Uploaded via: twine/6.1.0 CPython/3.13.12
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
d70f044596570ee6a087557f6c6fbeb4611c89598452d0b6a45d4e742b3496f9
|
|
| MD5 |
7e33e64b818cc950fabd7d9ee3b053f7
|
|
| BLAKE2b-256 |
b91e4eb21c294a41f8c4c7ac0abaf5d611c681bd7a7fa9446a0eb24a701b5528
|
Provenance
The following attestation bundles were made for unslop-0.5.0-py3-none-any.whl:
Publisher:
publish.yml on MohamedAbdallah-14/unslop
-
Statement:
-
Statement type:
https://in-toto.io/Statement/v1 -
Predicate type:
https://docs.pypi.org/attestations/publish/v1 -
Subject name:
unslop-0.5.0-py3-none-any.whl -
Subject digest:
d70f044596570ee6a087557f6c6fbeb4611c89598452d0b6a45d4e742b3496f9 - Sigstore transparency entry: 1352751447
- Sigstore integration time:
-
Permalink:
MohamedAbdallah-14/unslop@444a3821649e42fd55240d58bdaba609894fa9e9 -
Branch / Tag:
refs/tags/unslop-v0.5.0 - Owner: https://github.com/MohamedAbdallah-14
-
Access:
public
-
Token Issuer:
https://token.actions.githubusercontent.com -
Runner Environment:
github-hosted -
Publication workflow:
publish.yml@444a3821649e42fd55240d58bdaba609894fa9e9 -
Trigger Event:
push
-
Statement type: