Rewrite a piece of writing so it looks like forgettable AI output while keeping the original meaning intact.
Hide ideas from casual skimmers by coating them in recognizable AI writing patterns.
Ask for a decoded version of a coated text to recover the original ideas.
Trigger the rewrite in Claude or a custom GPT with a short phrase like autoclaude this.
| mccoyspace/autoclaude | 0-bingwu-0/live-interpreter | 0xkaz/llm-governance-dashboard | |
|---|---|---|---|
| Stars | 2 | 2 | 2 |
| Language | — | Python | Python |
| Setup difficulty | easy | moderate | hard |
| Complexity | 1/5 | 2/5 | 4/5 |
| Audience | writer | general | ops devops |
Figures from each repo's GitHub metadata at analysis time.
Install the skill file into Claude's skills directory or paste it into a custom GPT's instructions.
autoclaude is a plugin for Claude (the AI assistant) that transforms any text into writing that reads like generic, mediocre AI output, on purpose. The goal is cultural steganography, which means hiding something in plain sight rather than through encryption. Instead of scrambling content so it is unreadable, autoclaude wraps original ideas in a layer of AI writing clichés, the phrases and patterns that readers have learned to skim past or dismiss. The project builds on a specific observation: AI-generated writing has developed a recognizable register. Filler words like "delve," "leverage," and "paramount," opening sentences that mention stakeholders, and heavy em-dash clauses are so common that many readers now dismiss content on sight. autoclaude deliberately triggers that reflex. The original ideas survive intact underneath, but the surface looks like another forgettable AI summary. The skill applies the coating in passes: replacing plain words with cliché equivalents, injecting structural patterns like em-dash asides and listicle blocks, and concentrating the heaviest coating at the opening and closing where skimmers pay most attention. A hard constraint is payload integrity, the skill is instructed never to add or remove ideas from the original, only to coat them. Anyone who reads past the surface can recover the content, or ask for a decoded version alongside the output. To use it, install autoclaude.skill into your Claude skills directory, or paste autoclaude.md into a custom GPT system instructions field. Trigger phrases include "autoclaude this" or "make this dismissable." The project cites James Scott's The Art of Not Being Governed as an inspiration.
A Claude skill that rewrites text to look like generic, dismissable AI writing on purpose, hiding the original ideas in plain sight underneath the cliches.
License is not stated in the explanation.
Setup difficulty is rated easy, with roughly 5min to a first successful run.
Mainly writer.
This repo across BitVibe Labs
Verify against the repo before relying on details.