What does "SLSA Level 3" really mean? Securing an open-source build, live, with an AI agent

You have probably seen the claim: "Go from zero to SLSA Level 3 in minutes." But what does SLSA Level 3 even mean, and can you really do it on your own laptop?

In this talk we try it and find out. We take a normal open-source project and add real, signed proof to its build: where the code came from, what went into it (an SBOM), and whether it has any known vulnerabilities. Every step is signed as in-toto evidence, and a policy blocks the release if a critical CVE shows up.

The tool is cilock (open source, Apache 2.0). It is the next generation of Witness, a well-known supply-chain project that is part of in-toto, a CNCF graduated project. It was built by the Witness team to be driven by AI agents, not humans, so we let an AI coding agent run the commands for us. As we go, I explain every term in plain language: provenance, SBOM, in-toto, signing, SLSA. No prior knowledge needed.

Then the honest part: what you actually get on a laptop, versus what real SLSA Level 3 needs (signing keys kept away from the build, and a clean, isolated build machine), and how moving to a CI runner gets you the rest of the way. You will leave able to do all of this on your own project the same day.

1. What SLSA levels really mean, and what you can honestly claim on a laptop.

2. Plain-language intro to provenance, SBOMs, in-toto, and signing.

3. How the policy blocks a bad release and proves a binary was not tampered with.

4. How AI agents can run security tooling for you, with you still in control.