agent-readiness-scan

Agent-Readiness Scan (isitagentready.com)

Produce the official Cloudflare agent-readiness result for a domain as audit-grade artifacts: fixed-schema CSV + raw evidence + the 0-100 score.

Critical facts (learned the hard way)

1. Two sources, both required:
   • POST https://isitagentready.com/api/scan body {"url":"https://<domain>"} → full JSON (level, levelName, per-check status + embedded request/response evidence, nextLevel remediation prompts + skillUrls). No numeric score in the JSON.
   • The 0-100 score renders only in the web UI. Playwright-navigate to https://isitagentready.com/<host> and read the score dial: an <svg> carrying aria-label="Overall score: N out of 100". Never wait_for that text: it's an attribute, not visible text (text-waits time out). Take a page snapshot or evaluate document.querySelector('[aria-label*="Overall score"]'). A fresh "Last scanned" timestamp = results are rendered (the step-1 API POST itself refreshes the scan, so no Scan click is normally needed). Without the 0-100 you cannot track deltas (e.g. a site improving 21→43 after fixes).
2. Don't freeze the checklist. The scanner evolves (new checks appear). Emit whatever .checks returns, mapped through the fixed CSV schema; never hand-author check rows.
3. Scored vs supplementary. llms.txt, llms-full.txt, security.txt are NOT scored by Cloudflare but Theo audits track them. They go in Supplementary (not scored) rows from curl probes, never mixed into scored categories.
4. Commerce is informational unless isCommerce is true: one NOT CHECKED row, "does not affect score".

Workflow

scripts/run_scan.sh https://<domain> <outdir>/raw-data

Then get the official score via Playwright (see Critical fact 1 for the exact method). Save the page snapshot to raw-data/isitagentready_<client-slug>_snapshot.txt (client slug, e.g. acme, matching your audit config slug). Then:

python3 scripts/scan_to_csv.py \
  <outdir>/raw-data/iar_scan.json --score <N> --out <outdir>/csv-base-data/agent_readiness_checks.csv

(csv-base-data/ is the Theo full-pack convention; scoped snapshots have used plain csv/; either is fine, pass --out explicitly.)