Journeys
A guide answers one question. A journey takes you from a question to something you can show a colleague. Each one runs on a small repository written for the page, uses real output (every command block is run against the CLI before the site is published), and ends with what you proved: the edge or trust boundary you looked at, and how well.
- AppSec: find and prove a chain: Follow a model reply across two files into a shell, keep the proof as a verifiable bundle, break the edge. 25 minutes.
- Platform and CI: gate pull requests: Accept old findings in writing, compare against a baseline, publish the outputs and keep evidence. 30 minutes.
- AI engineer: the Trust Canvas: Draw an agent’s zones, tools and data, read the boundaries and chokepoints, and see what two approvals change. 35 minutes.
- Integrator: SARIF, CycloneDX, LSP, connectors: Take each output the way another system will, and find where data can leave the machine. 40 minutes.
What you need
Section titled “What you need”The Keryn CLI, git, and jq for the steps that read JSON. The AI engineer journey also uses the Trust Canvas in a browser. Nothing is uploaded: the scans and the canvas run on your machine, and the one step that can send findings to a model is a button you press and confirm.
Where the concepts live
Section titled “Where the concepts live”The journeys link to these instead of repeating them: the edge section for trust boundaries, edge kinds and chains, the rule catalog for what each rule inspects, and the taxonomies for how rules map to OWASP, CWE and NIST AI RMF entries.