v1.0.0 — Initial release

ai-txt skill v1.0.0

A Claude Code skill that generates and audits the files websites use to control or inform AI systems: robots.txt (with AI-specific user-agents), aitxt.ing-format /ai.txt, llms.txt, and /.well-known/tdmrep.json.

What's in this release

• Generate mode — picks the right file for the user's actual goal (block AI training, prevent hallucination, EU TDM opt-out, describe docs to LLMs).
• Audit mode — reviews existing robots.txt / ai.txt for deprecated tokens (anthropic-ai, claude-web), the Applebot-Extended inheritance trap, and missing modern AI user-agents.
• Bundled reference data (April 2026): current AI user-agent list, IP-range endpoints for spoof verification, ready-to-adapt file templates.
• Disambiguation — knows the difference between the three unrelated specs that all share the name "ai.txt" and routes the user to the right one.

Benchmark

On 3 realistic test cases (generate robots.txt, generate anti-hallucination ai.txt, audit a legacy robots.txt):

• With skill: 100% pass rate (25/25 assertions)
• Without skill baseline: 79.4% (20/25) — mis-identifies the right file format, blocks the wrong bots, misses the Applebot-Extended trap
• Delta: +21 percentage points

Install

```bash

Global

git clone https://github.com/First-Point/ai-txt-skill.git ~/.claude/skills/ai-txt

Or via the packaged .skill file (attached below)

unzip ai-txt.skill -d ~/.claude/skills/
```

Or add to skills-lock.json:
```json
{
"skills": {
"ai-txt": { "source": "First-Point/ai-txt-skill", "sourceType": "github" }
}
}
```

License

MIT © FIRST POINT YAZILIM LİMİTED ŞİRKETİ