ai-slop-detect 0.1.0 initial release

Free CLI to find LLM-generated text patterns. Pattern catalog: 43 English
phrases, 29 Polish phrases, 7 character tells (em-dash, curly quotes, etc),
13 decorative emoji, 7 weak openers. Deterministic, zero dependencies, MIT.

Built from real production use editing AI-drafted Polish FB posts and
English README files. Score bands: STRONG_AI (70+), LIKELY_AI (40-69),
TRACES (15-39), HUMAN_LIKE (<15). CI gate via --fail-on N.

Tests 5/5 pass.

Co-authored-by: Gniewislawa AI / Hermes <gniewka@antydizajn.pl>

Author: Paulina Janowska

.gitignore

@@ -0,0 +1,10 @@
+__pycache__/
+*.py[cod]
+*.egg-info/
+*.egg
+build/
+dist/
+.venv/
+.pytest_cache/
+.coverage
+.DS_Store

LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Paulina Janowska / Gniewislawa AI
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

README.md

@@ -0,0 +1,116 @@
+# ai-slop-detect
+
+Find LLM-generated text patterns in markdown, prose, blog posts, README files, and code comments. Free CLI, zero dependencies, deterministic. No API key required.
+
+Built from real production use editing AI-drafted Polish FB posts and English README files. Detects 70+ patterns across EN and PL, character-level typographic tells (em-dash, curly quotes, decorative emoji), and weak openers ChatGPT and Claude overuse.
+
+## install
+
+```bash
+pip install ai-slop-detect
+```
+
+Or run from source:
+
+```bash
+git clone https://github.com/antydizajn/ai-slop-detect
+cd ai-slop-detect
+python -m ai_slop_detect README.md
+```
+
+## use
+
+Scan a single file:
+
+```bash
+ai-slop README.md
+```
+
+Scan a directory (defaults to md/markdown/mdx/txt/html/rst):
+
+```bash
+ai-slop docs/
+```
+
+Pipe text from stdin:
+
+```bash
+cat draft.md | ai-slop -
+```
+
+JSON output for tooling:
+
+```bash
+ai-slop docs/ --json > slop_report.json
+```
+
+CI gate — exit 2 if any file scores 50 or higher:
+
+```bash
+ai-slop docs/ --fail-on 50 --quiet
+```
+
+## output
+
+```
+== README.md ==
+verdict: LIKELY_AI (ai_score=47/100)
+stats: 12 hits across 320 words (3.75 per 100w)
+by_kind: char:em_dash=4, opener=2, phrase=6
+
+hits:
+     3:18  [phrase] 'in conclusion'  | In conclusion, this is a great project.
+     5:42  [char:em_dash] '\u2014'   | This tool — built from scratch — finds AI tells.
+    12:1   [opener] 'Certainly,'     | Certainly, here is the implementation:
+```
+
+Score bands:
+
+- `STRONG_AI` (70+): unmistakably AI, rewrite from scratch
+- `LIKELY_AI` (40-69): heavy editing needed
+- `TRACES` (15-39): polish remaining tells
+- `HUMAN_LIKE` (<15): ship it
+
+## what it catches
+
+**English phrases** (43 patterns): "let me know if", "as an AI", "in the realm of", "delve into", "harness the power", "cutting-edge", "transform your workflow", "in today's digital landscape", and more.
+
+**Polish phrases** (29 patterns): "ponadto", "warto zauważyć", "rewolucyjny", "kluczowym aspektem", "mam nadzieję, że ten post", and more.
+
+**Character tells**: em-dash (—), en-dash (–), curly quotes (" " '), horizontal ellipsis (…), non-breaking spaces.
+
+**Decorative emoji**: 🚀 ✨ 💡 💪 🎉 🔥 🌟 🎊 🌈 💫 ⭐ 🎯 💯 (the marketing-AI set).
+
+**Weak openers**: "Certainly,", "Of course,", "Absolutely,", "Great question,", "I'd be happy to", "That's a fantastic".
+
+## why
+
+LLM text has a fingerprint. Em-dash density, "delve" frequency, opener cadence, decorative emoji clusters. If you publish text that scores LIKELY_AI on this tool, readers (and other tools) can tell. This linter gives you a fast pre-publish gate.
+
+Built because Paulina, the author, needed it for her own Polish FB posts (Gniewislawa AI / Hermes). The PL pattern catalog is tested against real production drafts.
+
+## pricing
+
+The CLI is free and MIT-licensed. If you want **AI-powered rewrite suggestions** (BYO Claude/OpenAI key) or **bulk repo scan with HTML report**, those land as a paid premium tier later. Free tier covers 100% of detection.
+
+If this tool saves you 30 minutes of editing, consider tipping:
+
+- BLIK / Polish bank: `793 093 721`
+- Revolut: `@paulinajeoo` (or `revolut.me/paulinajeoo`)
+- PayPal: coming soon
+
+## roadmap
+
+- HTML report with line highlighting
+- Per-language profile flags (`--lang en`, `--lang pl`, `--lang both`)
+- Pre-commit hook template
+- GitHub Action wrapper
+- Configurable custom patterns via `.aisloprc.toml`
+
+## contributing
+
+Pattern PRs welcome. New phrase? Open an issue with three real examples and the source LLM. We add patterns that catch real AI text without flagging human writing.
+
+## license
+
+MIT