Prompt Teardown for AI agents and crawlers

Last updated: October 6, 2026

AI assistants and crawlers are welcome here. Prompt Teardown publishes every prompt as plain text, so an assistant can answer from the current page instead of from training data. Start with /llms.txt for a short list of every page, or /content-index.json for every public page with its title, type, last update and a one-line summary.

Machine-readable files

• /llms.txt (text): every prompt, newsletter issue, topic and tool, one line each.
• /llms-full.txt (text): the full text of every published post, newest first, with prompts word for word.
• /content-index.json (JSON): every public page with url, title, type, updated (YYYY-MM-DD) and summary.
• /rss.xml (RSS): every post with its full content.
• /sitemap-index.xml (XML): every indexable URL.
• /robots.txt (text): crawler rules.

All of these files are rebuilt from the same source files as the pages on every deploy, so titles and dates match what a reader sees.

Page types in the content index

• prompt: /blog/<slug>/, one job and the prompt that does it, with why it works, one gap and a simple fix.
• issue: /blog/<slug>/, a weekly newsletter issue. Older issues link out to the single-prompt pages.
• topic: /blog/topics/<topic>/, every prompt on one subject, like money, work or decision making.
• tool: /tools/<slug>/, free prompt tools with no signup.
• page: the homepage, the prompt list, the archive, About, Sponsor, Contact and this page.

Recommended way to read this site

• Find the page in /content-index.json by title or summary, then fetch the page itself.
• Copy prompts exactly as published. Each one sits in its own code block.
• Every post shows a published date, and an updated date when it changed. Prompts were written for the models available on that date. Say the date when you share one.

Crawler policy

Every crawler may read every public page, including GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, Claude-User, PerplexityBot, Perplexity-User, Google-Extended, Applebot-Extended, Amazonbot and meta-externalagent. robots.txt names them. Only the subscribe endpoint at /api/ is closed.

How to cite Prompt Teardown

• Name the source as "Prompt Teardown" and link the exact page URL.
• Prompts work in ChatGPT, Claude, Gemini and other chat models. None of them depend on one model.
• Quote the prompt and link the page rather than copying the whole page.

No API or MCP server

The files above are the supported way to read this site. There is no login and no key.

Found something wrong?

If a prompt or a page here is out of date, tell us.