slop-grader
lukstei · Writing · Marketing · AI Detection
slop-grader is a command-line tool that grades text files against rules you write yourself, then flags the lines that break them. As an AI slop detector, it's built for anyone who wants to catch AI writing habits like banned words, empty adverbs, and rhetorical formulas before publishing. The tool runs every rule against every line in parallel through Jev, a TypeSafe AI model that returns probabilities instead of generated text. A full check costs a fraction of a cent.
About slop-grader
What Is slop-grader
slop-grader is an open-source CLI published on npm as @lukstei/slop-grader. You point it at a file, hand it one or more rulesets, and it returns a report listing the lines that crossed a confidence threshold. The rules are plain Markdown, so you describe what "slop" means for your case in your own words.
Most AI text checkers send a whole document to a generative model and hope it doesn't skip anything. slop-grader splits the work. Line rules test individual sentences for specific patterns, while document rules rate the text as a whole on things like narrative arc and closing strength. Because each check is a yes/no question answered by a System One model, this rule-based text grader can run thousands of them without paying for generated tokens.
The catch is that it's a developer tool at heart. You need Node.js 18 or newer, a terminal, and an API key from either TypeSafe AI or OpenRouter. There's no web dashboard. Nothing works offline once you actually grade a file.
Getting Started
- Install Node.js 18 or newer if you don't have it, then grab an API key from TypeSafe AI or OpenRouter.
- Set the key in your shell, for example
export TYPESAFE_API_KEY=.... - Pick your rulesets and run the grader, for example
npx @lukstei/slop-grader@latest -r no-ai-slop my-draft.md. - Read the report, then paste it into your AI coding agent so it can plan fixes and produce a cleaned-up version.
- Re-run the command on the edited file. Unchanged lines resolve from cache, so only the parts you touched cost anything.
Product Information
A quick look at slop-grader's pricing, supported platforms, and performance.
Best for
The users, tasks, and scenarios where this tool fits best.
Users
- Writers and editors
- Developers writing docs
- Content and marketing teams
- QA and reviewer engineers
Tasks
- Catching AI writing habits
- Scoring document quality
- Auditing technical documentation
- Reviewing code diffs
- Checking non-prose files
Scenarios
- Final pass before publishing a blog post
- Reviewing a teammate's pull request
- Building a repeating editorial loop
- Scanning large files on a budget
Key features
Parallel Line and Document Grading
slop-grader separates line rules from document rules. Line rules test one sentence at a time against a specific pattern, while document rules rate the whole text on softer qualities like tone and arc. Both run through the same pipeline, and clean lines are simply left out of the report so you only see what failed.
Plain Markdown Rulesets
You write rules as Markdown files, not code. A rule is a plain-text question plus criteria for what counts as true or false, which means a non-programmer can draft one and a developer can drop it into the run. Built-in rulesets like no-ai-slop, article-scores, tech-docs, and grammar-english cover common cases out of the box. The no-ai-slop ruleset is the quickest way to detect AI writing in a fresh draft, while custom rulesets let you target your own phrasing.
Pay-Per-Check Cost Model
There's no subscription. You pay only for the questions the tool actually asks, billed through your TypeSafe AI or OpenRouter key. This text grading CLI prices by the check itself. Checking a 16-line sample against 53 rules runs about half a cent, and a 2,100-word article against 42 rules lands near 7.6 cents.
Incremental Caching
Every line gets cached by a hash of its content and the rule's criteria. Edit a document and re-run. The untouched lines come back from cache with zero API calls. This makes the edit-fix-repeat loop fast and cheap on long files, and you can disable it with --no-cache when you want a clean sweep.
Structured JSON Output
Swap the human-readable report for JSON with the --json flag. Violations come out as line numbers, offending text, and the rule names that fired, plus per-document scores with confidence values. That shape is meant to feed CI pipelines and editor integrations rather than a person reading a terminal.
Provider Flexibility
The tool talks to two backends: TypeSafe AI for direct Jev access and OpenRouter. It auto-detects which one to use from your environment variables, or you can force it with --provider. The default model is jev-1.13.0, and you can override it per run.
Offline Rule Validation
A --check flag validates your ruleset syntax without grading anything or spending a cent. Write a new ruleset and check it first. That way you know it parses before you point it at real text.
Pros and cons
Pros
- Cheap per-run cost, since grading returns probabilities instead of generating long text.
- Custom rules are readable Markdown, so you can tune them without touching code.
- Caching keeps repeated runs on big files fast and nearly free.
- JSON output slots straight into CI and editor tooling.
- Open source under MIT, so you can read or fork the whole thing.
Cons
- No GUI or web app. Everything happens in a terminal, which rules out non-technical users.
- You must supply your own API key from TypeSafe AI or OpenRouter, so there's no all-in-one hosted option.
- It judges text, not facts. The tool can't tell you whether a claim is accurate, only whether the wording looks like slop.
- The project is young, and the earliest releases landed only days before this writing.
Frequently asked questions
It reads a text file, tests every line against the rules you give it, and prints a report of the lines that look like slop. You define what counts as slop through Markdown rulesets, so the tool adapts to your definition.
Related content
Explore related tools, skills, and articles for slop-grader.
slop-grader Alternatives
Emailwritr
Emailwritr · Writing · MarketingEmailwritr is an AI email copywriting tool built for affiliates, course creators, and online businesses that want more revenue from the list they already own. Instead of blank pages or generic chatbot prompts, this AI email writer generates welcome sequences, launch campaigns, cart recovery emails, and promo series from frameworks modeled on well-known direct-response copywriters. You connect your autoresponder, answer a few questions about your offer, and it writes the copy for you.
Swms
Swms AI · Writing · Productivity · BusinessSwms is an AI safety compliance tool that turns a plain description of a job into ready-to-use safety documents. Describe your project, trade, and known hazards, and it writes a safe work method statement, job hazard analysis, safe work procedure, RAMS, or safety data sheet in seconds. It also ships with Oscar, a chat assistant that answers workplace safety questions in any language. Teams in construction, mining, transport, and warehousing use it to cut the paperwork that usually eats into a workday.
Prosp
Prosp · Writing · Voice & Language · MarketingProsp is an AI LinkedIn outreach tool built for agencies and sales teams, and it writes the message and the voice note in your own voice for each prospect so they actually reply. You connect your accounts, find leads, and let the AI draft and send personalized messages at scale, all from one inbox. It's built for people running outbound at volume. That's the whole pitch. Every touchpoint still has to feel human.