Official CAIN-42 IHVRK (Independent Human Verification Resource Kit): Skill Memory Sandbox
Don't trust CAIN-42. Verify CAIN-42.

What the product is said to do:
  The Skill Memory Sandbox encrypts each AI skill's memory under its own key, so one skill (say, a smart-home add-on) cannot read another's secrets (say, a password vault), even when it runs in the same agent.

This kit holds 4 claim(s) from https://cainstudio.online/guide-skill-sandbox.html, each as the exact request and the value its answer
must contain. They were run live on 2026-11-03T22:54:53+00:00 and every answer matched (a run that does not match publishes no kit).

Check it yourself (Python 3.8+, standard library only, no CAIN-42 code, no account):
  python3 verify.py             offline: hashes, Ed25519 signature, recorded answers vs. claims
  python3 verify.py --live      also send every request again from YOUR machine and check the answers now

Compare the key fingerprint verify.py prints with https://cainstudio.online/evidence-kit-key.json,
https://mcpgate.online/evidence-kit-key.json and https://clawx.click/evidence-kit-key.json.

Scope: these claims show what the live demo does with these inputs. They are not a third-party audit, a
certification or a guarantee for inputs not listed. See the product's limits page for what it does not do.
Files: README.txt, verify.py, claims.json, transcript.json, kit-key.json, MANIFEST.json, MANIFEST.sig
