# Text > Use this skill when the user wants to create, generate, save, export, download, convert, format, clean, or localize plain text, TXT, Markdown/MD, or HTML file deliverables. This skill produces text-family artifacts such as .txt, .text, .md, .markdown, or .html files and exposes the final file for download. - Skill: `cloudchef/text` (Agent Skill, multi-file: 3 files) - Install (CLI): `npx skillmds@latest add cloudchef/text` - Raw SKILL.md: https://api.skillmd.com/api/skills/cloudchef/text/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Web & Frontend - License: MIT-0 - Author: CloudChef (https://skillmd.com/u/cloudchef) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/cloudchef/text --- # AtlasClaw Workspace Output When producing a user-facing `.txt`, `.md`, `.markdown`, or `.html` file: 1. Write only the final content under the current work_dir, preferably `exports/.` unless the user requested a safe relative filename. 2. Use UTF-8 text. For HTML, produce a complete standalone document when the user asks for an HTML file. 3. Expose the final file through the existing workspace download flow by passing its work_dir-relative path as `download_paths`. 4. Prefer writing a new output file instead of editing source files in place unless the user explicitly asks for an in-place edit. 5. Do not expose scripts, logs, caches, hidden files, intermediate files, or source documents as downloads. ## AtlasClaw Runtime Commands Prefer `skill_read`, `skill_write`, and `skill_edit` for simple file operations. If a transformation needs shell features such as pipes, redirects, globs, or compound commands, run it through `bash -lc ''` with explicitly named work_dir-relative files, or use a `python3 -c` command that reads and writes explicit paths. Do not read or print `.env`, key, token, cookie, or password files unless the user explicitly requests it and the output is redacted. ## Quick Reference | Task | Load | |------|------| | Text cleanup, extraction, Markdown, or HTML shaping | `data.md` | | Translation/localization for text output | `localization.md` | --- ## Universal Text Rules ### Encoding - **Always verify encoding first:** `file -bi document.txt` - **Normalize line endings:** `tr -d '\r'` - **Remove BOM if present:** write a normalized copy instead of editing the source in place. ### Whitespace - **Collapse multiple spaces:** `sed 's/[[:space:]]\+/ /g'` - **Trim leading/trailing:** `sed 's/^[[:space:]]*//;s/[[:space:]]*$//'` ### Common Traps - **Smart quotes** (`\u201c`, `\u201d`, `\u2018`, `\u2019`) break parsers -> normalize to ASCII quotes. - **Em/en dashes** (`\u2013`, `\u2014`) break ASCII-only output -> normalize to `-`. - **Zero-width chars** invisible but break comparisons → strip them - **String length ≠ byte length** in UTF-8 (`"café"` = 4 chars, 5 bytes) --- ## Format Detection ```bash # Detect encoding file -I document.txt # Detect line endings cat -A document.txt | head -1 # ^M at end = Windows (CRLF) # No ^M = Unix (LF) # Detect likely HTML grep -qiE '"{}|\\^`\[\]]+'` | --- ## Before Processing Checklist - [ ] Encoding verified (UTF-8?) - [ ] Line endings normalized - [ ] Target format/style defined (TXT, Markdown, or HTML) - [ ] Edge cases considered (empty, Unicode, special chars)