Benchmark Your Site with the Agent Readiness Scan Tool

The agent-readiness-scan skill gives your GEO audit an independent external benchmark. It connects to Cloudflare’s public isitagentready.com scanner, runs a corroboration matrix of curl probes alongside the official API, captures the 0 to 100 score from the web UI via Playwright, and converts everything into a fixed-schema CSV you can track across re-audits. The delta between audit runs (not the first number) is the proof that your fixes are working.

The 0 to 100 score is only available in the isitagentready.com web UI, rendered in an SVG dial. The JSON API returns detailed per-check findings but not the composite score. The Playwright capture step is not optional if you want to track deltas.

How It Works

The scan runs in two scripts that work together:

run_scan.sh: Official API + Corroboration Matrix

run_scan.sh POSTs to the isitagentready.com API and then runs a series of curl probes to corroborate the findings independently:

bash skills/agent-readiness-scan/scripts/run_scan.sh https://example.com ./raw-data

What the script captures:

Probe	What it checks
`POST https://isitagentready.com/api/scan`	Official scored results, level, per-check status, and embedded request/response evidence
HTTP headers	Homepage response headers including content-type and caching signals
Markdown negotiation	Whether the server responds to `Accept: text/markdown` requests
`robots.txt`	Full content for AI-crawler access analysis
`llms.txt`	Present / absent (supplementary, not scored by Cloudflare)
`security.txt`	Present / absent (supplementary, not scored)
MCP endpoint	`/.well-known/mcp.json` detection
OAuth discovery	`/.well-known/oauth-authorization-server` detection

The script writes all raw artefacts to the output directory: iar_scan.json, homepage_headers.txt, accept_markdown_headers.txt, robots.txt, supplementary.json, and more.

Playwright: Capture the 0 to 100 Score

After the API scan completes, navigate to https://isitagentready.com/<your-domain> in a Playwright-controlled browser and read the score from the SVG dial:

// The score is an aria-label attribute, not visible text.
// Text-based waits time out. Use the attribute selector instead:
document.querySelector('[aria-label*="Overall score"]')

A fresh “Last scanned” timestamp on the page confirms the results are rendered. Save the page snapshot to your output directory as isitagentready_<client-slug>_snapshot.txt.

scan_to_csv.py: Convert to Fixed-Schema CSV

Pass the scan JSON and the captured score to scan_to_csv.py:

python3 skills/agent-readiness-scan/scripts/scan_to_csv.py \
  ./raw-data/iar_scan.json --score 72 --out ./csv-base-data/agent_readiness_checks.csv

The script maps Cloudflare’s check keys to fixed display names and emits a CSV with consistent category headers across every client and every re-audit.

Using from Claude Code

Ask for it in plain language from your Claude Code session:

run an agent-readiness scan on example.com

Or run the scripts manually from your terminal:

# Step 1: Run the API scan + corroboration matrix
bash skills/agent-readiness-scan/scripts/run_scan.sh https://example.com ./raw-data

# Step 2: Get the 0-100 score via Playwright (navigate to the URL, read the SVG dial)
# Save the score as a number, e.g. 72

# Step 3: Convert to fixed-schema CSV
python3 skills/agent-readiness-scan/scripts/scan_to_csv.py \
  ./raw-data/iar_scan.json --score 72 --out ./scan.csv

Output CSV Schema

The CSV uses a fixed schema so results stay comparable across clients and re-audits:

Column	Contents
`category`	Category name with pass/fail tally, e.g. `Discoverability (3/4)`
`check`	Individual check name, e.g. `robots.txt`, `DNS for AI Discovery (DNS-AID)`
`result`	`PASS`, `FAIL`, or `NOT CHECKED`
`detail`	The scanner’s message verbatim
`source`	Data source and date, e.g. `isitagentready.com + curl 2026-06-14`

The first data row is always the OVERALL row:

OVERALL, Agent-readiness score, 72/100 - Level 2 "Discoverable", 8 pass / 3 fails; Commerce not applicable, isitagentready.com/example.com (Cloudflare) 2026-06-14

Scored categories (from the Cloudflare scanner):

Discoverability
Content Accessibility
Bot Access Control
API Auth MCP & Skill Discovery
Commerce (marked NOT CHECKED unless isCommerce is true)

Supplementary rows (from run_scan.sh curl probes, not scored by Cloudflare):

llms.txt, PRESENT or ABSENT
llms-full.txt, PRESENT or ABSENT
/.well-known/security.txt, PRESENT or ABSENT

Never mix supplementary checks into the scored category tallies. Cloudflare does not score llms.txt, tracking it separately preserves the accuracy of your scored pass/fail counts.

Common Mistakes to Avoid

Reporting only the Level, not the 0 to 100 score

The Cloudflare level name (e.g., “Discoverable”) does not encode the numeric score precisely enough to track deltas. Always capture the 0 to 100 value from the web UI dial via Playwright. A site improving from 21 to 43 after fixes is the evidence your client wants to see.

Using WebFetch to scrape isitagentready.com

The site is a JavaScript app, a WebFetch or plain curl returns only an empty shell with no results. Use the official POST https://isitagentready.com/api/scan API for check data, and Playwright for the score dial.

Renaming or inventing category names

Cross-client comparability depends on exact category labels. Use the fixed names in scan_to_csv.py (e.g., API Auth MCP & Skill Discovery, not Protocol Discovery). The scanner adds new checks over time, emit whatever .checks returns, mapped through the fixed schema.

Placing This in the THL Method

The agent-readiness scan is the second layer of the three-layer THL GEO Method, sitting between the dimensional geo-audit and the client deliverable from audit-report-kit. Its purpose is specifically to provide a score that your own dimensional audit cannot give itself, an independent external benchmark from Cloudflare’s infrastructure. Record both scores (your dimensional composite and the Cloudflare 0 to 100), and on every re-audit, track the delta on each.

API Reference

MCP Server

Open-Source Skills

Benchmark Your Site with the Agent Readiness Scan Tool

How It Works

Using from Claude Code

Output CSV Schema

Common Mistakes to Avoid

Placing This in the THL Method

​How It Works

​Using from Claude Code

​Output CSV Schema

​Common Mistakes to Avoid

​Placing This in the THL Method

How It Works

Using from Claude Code

Output CSV Schema

Common Mistakes to Avoid

Placing This in the THL Method