Courtesy DM — Cloudflare report
Pick one recipient (LinkedIn first, then X, then email). Send before publishing the public X thread / LinkedIn post — gives the recipient a chance to react in private first, which builds the relationship better than reading about the score in their feed.
DM body (Josh’s email voice)
Howdy [Name],
We just shipped an Agent Readiness Report on Cloudflare against the Agents First rubric (the one your April Agent Readiness Score post is the canonical source for). www.cloudflare.com scored 85/100, Level 3 — top-tier in the series. Real respect: you wrote the rubric, scored yourself, and shipped the artifacts on the marketing root. That sequence is the credibility move.
The variance is the story though — developers.cloudflare.com scored 35/100 (Level 2), and blog.cloudflare.com scored 0/100 (Level 0, no agent access at all). Same engineering culture, demonstrably the same playbook one subdomain over. Closing the dev-portal gap is a flat-file edit (copy /AGENTS.md, /agents.json, named-bot robots.txt over) — ~30 min for +50 points on the surface that arguably matters more. The blog is the bigger surprise — whatever's blocking agents there is presumably an artifact of a different team's robots.txt defaults, not intent.
Full report: https://agentsfirst.dev/reports/cloudflare/
No need to respond — sharing in case useful, and gladly re-score once anything ships.
Yeehaw!
🤖 Josh
Recipient candidates
| Person | Role | Why them | Channel |
|---|---|---|---|
| Brendan Irvine-Broque | Workers / Developer Platform PM (Cloudflare) | Most likely to be the operator who decides what ships on developers.cloudflare.com next; PM tier where the marketing-root → dev-portal hand-off lives. Direct fix-path conversation. | LinkedIn DM — verify handle first |
| Matthew Prince | CEO, Cloudflare | Active on X, will see + likely reshare; relationship-building tier rather than fix-path tier. Best for the public-credibility win. | LinkedIn DM (linkedin.com/in/eastdakota) — verify before sending; or X (@eastdakota) |
| John Graham-Cumming | CTO, Cloudflare | Long-running tech blogger; would respect the rubric and the credibility frame. Also the natural owner-adjacent person for the blog.cloudflare.com 0/100 finding. | LinkedIn DM — verify handle first |
| Sam Rhea | Product, Cloudflare (frequently bylined on Workers/agents posts) | Often co-authors the developer platform launches; possible original ARS post co-author. Direct dev-portal-fix path if he’s the right person. | LinkedIn DM — verify handle first |
Pick rule: Brendan Irvine-Broque first — fix-path matters more than reach for the first send. Matt Prince is the natural follow-up if Brendan doesn’t respond within 5 days.
Channel order: LinkedIn DM > X DM > email (no public CF press address known to be agent-monitored).
Fallback timing: if no response within 5 business days from the first send, retry on the next-tier candidate. Don’t double-send to the same person.
If Cloudflare replies:
- “We just shipped the fix to developers.cloudflare.com” → re-run
/agentsfirst developers.cloudflare.com, post a follow-up X thread + comment on the original report (“CF moved L2→L3 in 6 days — here’s what landed”). The follow-up is better content than the original report. - “The blog 0/100 is a robots.txt default we didn’t realize was there” → great, even simpler fix; same follow-up pattern, and the “we didn’t realize” framing is itself a useful story about how blocking-by-default propagates.
- “Disagree with X” → engage; if their critique is legitimate, file as a
bugs/incoming/report against agentsfirst itself and update the rubric. - Silence → fine; the public post still lands, the relationship is on file for the next cycle.
Notes for the sender
- Verify handles before tagging anyone in the public posts. Per the global rule. The /grok-twitter –verify mode resolves the handle + confidence; don’t trust training-data recall.
- Send the DM 12–24h before the public post lands, not after. The recipient hearing about the score from you privately first is the difference between “courtesy” and “ambush.”
- Lead with the score, not the critique. 85/100 is real respect; the dev-portal and blog observations are constructive, not snarky.
- Don’t oversell the fix. “30 min for +50 points” is the honest characterization of a flat-file edit if the artifacts are repo-templated; if they aren’t yet, the real cost is “however long it takes to set up the template repo.” Either is fine; just don’t promise instant.
- The blog 0/100 is the spicy finding — handle it gently. It’s a worse score than most companies in the series get on any surface, and it’s on Cloudflare, the company that wrote the rubric. Frame as “presumably an artifact, not intent” in the DM; let the public post note the irony more directly if it needs to.