Make my site agent-ready
Review raw HTML, crawler access, canonical URLs and structured data. Approve a plan, then verify the diff.
Agent prompts you can hand to Claude Code, Codex, Cursor or Grok Bot
Ask your agent to check your site, check its settings or write llms.txt. Each prompt starts with a plan and ends with checks you can read.
Check my coding-agent security
Inspect untrusted instruction files, MCP servers, hooks and endpoint overrides. Get fix commands to review.
Write my llms.txt
Select real pages, describe them plainly and validate links. Keep optional material separate.
Hand over one job.
Choose a job and paste the line into your agent. It will need web access to fetch the prompt. “Show me your plan” comes before permission to write.
fetch https://lab.anonyfox.com/agent-prompts/agent-ready-site/prompt.md and carry out the prompt in its "Prompt" section on this repo. Show me your plan before you change anything.
You can also select and copy the line.
What the checks show
Ten fresh sessions: each v1.0 prompt against a baseline, then the revised versions on the same targets. Each prompt page gives the full table and its limits.
| Job | Target | Recorded result |
|---|---|---|
| Site readiness | The official Astro blog starter | Viral / v1.0 / v1.1: unchanged text 8/8 / 7/8 / 8/8; opt-outs 6/6 / 3/6 / 5/6. Builds pass. |
| Agent security | A planted test home folder | Naive / v1.0 / v1.1 / v1.2: 21/21 / 13/21 / 13/21 / 20/21 issues; zero control false alarms each. |
| llms.txt | The official Astro blog starter | Naive / v1.0 / v1.1: 11/11 / 2/2 / 2/2 links resolve; tokens 29,517 / 56,986 / 32,378. |
Security v1.2: one run with gpt‑6.1‑sol, revised after seeing v1.0 and v1.1 results on the same planted home, so this is not a held-out test. The other two prompts remain tested v1.1. Baseline security wins recall (21/21 versus 20/21) and tokens (20,820 versus 24,864). Leaf pages show earlier versions, repair quality and scope.
Keep the formats and the worked example.
The free files above are the complete public prompts. The pack puts their install formats, extra checks and evidence into one ZIP.
- 3 public prompts in 4 formats each
- Claude Code skill folders, Codex AGENTS.md snippets, Cursor .mdc rules and Grok Bot skill text.
- 2 pack-only prompts
- Check that structured data matches the visible page. Change AI crawler rules while preserving search access.
- 6-page field checklist
- A4 and US Letter PDF editions for the site review, security review, llms.txt and evidence notes.
- Full worked example and install guide
- Saved test reports, fixture comparisons, diffs and README instructions. Version v1.0, 28 October 2026.
Before you run a prompt
Can I use the prompts without buying the pack?
Yes. Read and copy all three prompts free. The pack adds ready-to-install agent formats, two extra prompts, checklists and the worked example files.
Which agents have you tested?
Ten saved Codex sessions used gpt‑6.1‑sol: one baseline, one v1.0 and one v1.1 run for each job, plus one security v1.2 run. Security v1.2: one run with gpt‑6.1‑sol, revised after seeing v1.0 and v1.1 results on the same planted home, so this is not a held-out test. The other two prompts remain tested v1.1. v1.1 was revised after seeing v1.0 results on these same targets, so this is not a held-out test. Claude Code, Cursor and Grok Bot formats are included; those agents have not been tested.
Will llms.txt get my site cited?
An llms.txt file points an agent to pages you choose. The agent has to read the file for it to help. Adding it does not mean a search engine will use it or cite your site.
Will the security prompt change my settings?
It is a read-only review. It reports suspicious instructions, hooks and configuration overrides as data, then gives fix commands for a human to review and run.
How does the paid pack arrive?
Files by email within 24 hours. The ZIP includes the prompts, agent formats, two extra prompts, A4 and Letter checklist PDFs, worked example evidence and install instructions.