# Brainrotify > Rewrite developer-facing prose into playful Gen-Z "brainrot" language without changing executable code. Use when an AI coding agent or assistant needs to transform comments, docstrings, Markdown, README files, or other documentation text surfaces in one file or a directory while preserving syntax, logic, formatting, and code fences. Support optional backup creation and three intensity levels: `L`, `Mid`, and `W`. - Skill: `tengfone/brainrotify` (Agent Skill, multi-file: 5 files) - Install (CLI): `npx skillmds@latest add tengfone/brainrotify` - Raw SKILL.md: https://api.skillmd.com/api/skills/tengfone/brainrotify/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: AI & ML - Author: tengfone (https://skillmd.com/u/tengfone) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/tengfone/brainrotify --- # Brainrotify Rewrite eligible text surfaces into meme-heavy prose while leaving program structure intact. Prefer the bundled script for file-based work because it handles file discovery, safe backups, comment targeting, and reporting more reliably than ad hoc edits. This skill is agent-agnostic. Use the same workflow with Codex, Claude, or similar AI coding agents that can read files, run a script, and present a diff or summary. ## Workflow 1. Initialize the skill clearly. In the first user-facing update, explicitly say that `brainrotify` runs locally for normal rewrites, list the available intensity levels (`L`, `Mid`, and `W`), and present the scope options up front: single file, folder, or preview-only. Recommended opening prompt: `brainrotify` runs locally for normal rewrites. Available intensity levels are `L`, `Mid`, and `W`, and I will use `Mid` by default unless you want a quieter or louder tone. I can brainrotify a single file, a folder, or do a preview-only pass first. What path do you want me to target, and do you want a timestamped backup in `.brainrot-backup/` before I write changes? 2. Confirm scope. Identify whether the user wants a single file, a folder, or a preview-only pass. Default to `Mid` unless the user asks for a quieter or louder tone. 3. Offer a backup before writes. Ask once whether to create a timestamped backup in `.brainrot-backup//`. Do not skip this question unless the user already said `no backup`, `overwrite`, or `preview only`. 4. Run the helper script. Use the script for real file rewrites. Prefer `--dry-run --show-diff` first for large batches or unfamiliar file types. ```bash python3 /scripts/brainrotify.py README.md --intensity L --dry-run --show-diff python3 /scripts/brainrotify.py docs/ src/ --intensity Mid --backup python3 /scripts/brainrotify.py . --intensity W --no-backup ``` 5. Review the result. Report how many files were brainrotified, how many were unchanged, and how many were skipped. Mention the backup location when one was created. ## Safety Rules - Rewrite comments, Markdown, README files, HTML comments, and real docstrings only. - Do not rewrite fenced code blocks, inline code, imports, identifiers, signatures, executable statements, or non-docstring string literals. - Keep headings, list structure, indentation, comment markers, and surrounding formatting intact. - Use `--dry-run --show-diff` before bulk writes when the support matrix is only partially relevant to the target tree. - Fall back to manual inspection when the file type is unsupported or the diff looks risky. ## Script Notes The script supports: - Markdown and plain text docs outside fenced code blocks and frontmatter - Python `#` comments and true docstrings - `//` and `/* */` comments in common JS/TS/C-family files - `#` comments in shell and config files when they are outside quoted strings - `` comments in HTML-like files - A separate lexicon file at `scripts/brainrot_lexicon.json`, so vocabulary updates do not require Python edits The script defaults to `Mid` intensity. In non-interactive mode it refuses to rewrite unless `--backup` or `--no-backup` is provided explicitly, which prevents silent destructive runs. Keep runtime rewrites network-free. If you want fresher slang, update the lexicon file offline or add a separate refresh script that fetches and compacts terms into the local JSON cache. Do not make live scraping a requirement for normal rewrites. ## Intensity Guide - `L`: keep the text readable; add a few slang substitutions and light emphasis - `Mid`: push harder on meme phrasing while keeping the message easy to follow - `W`: lean into chat-brained phrasing, but keep the meaning recoverable Use [tone-guide.md](references/tone-guide.md) when you need to reason about style choices or explain the intensity differences before rewriting.