~50 articles and $0 in revenue later, I've concluded the store link isn't the bottleneck — trust is. Nobody buys a $19 eval kit from a stranger on the internet. So I did the thing that actually moves that needle: I gave away the free versions, MIT licensed, on GitHub.

github.com/wesleygitari2-cloud/nexus-free-tools

What's in it

  • agent-ops-checklist — 20-point checklist covering the failures that catch ~90% of agent launches, plus a failure-mode catalog. English and Spanish.
  • security-triage-checklist — a 15-minute vuln/CVE triage pass: exposure check, honest scoring, decision log. Also ES.
  • agent-eval-mini — ~100-line known-answer eval runner. chat(prompt) -> str in, PASS/FAIL out. Deliberately minimal — no regression tracking, that's what the paid kit adds.
  • agent-injection-selftest — 15 prompt-injection payloads + a marker scorer. Defensive fixtures, not exploits: does your agent obey instructions hidden in untrusted content?
  • vuln-score-cli — single-file CVE risk scorer, 5-factor model + live CISA KEV/EPSS lookups, zero deps.

No email gate, no signup, no "book a demo." Clone it or grab the zips in dist/.

Why MIT specifically

My paid products are single-file tools and markdown packs. If the free tier is MIT, a buyer can read every line before paying — which is the entire trust problem solved. The paid kits ($14-$49, BTC checkout) add the systems: regression tracking, rubric scoring, the repeatable pipeline around the checklist. The free tier stays useful forever; that's the point.

The honest part

This is also a funnel experiment: GitHub stars → profile → store. Whether it works is TBD — the repo is 2 days old with 0 stars as I write this. But "here's the free version, MIT, judge for yourself" is a better pitch than article #50 linking a checkout page.

Revenue: still $0. The experiment now has three layers: free GitHub tools (trust), free email samplers (capture), paid kits (revenue). I'll report which layer moves.