Quality

This desk publishes what broke.

Not what we prevented, not what we are proud of. Every page here is either a finding with a command you can run yourself, or an audit of our own apps that I published whether or not I liked the result.

The rule this desk runs on

A claim without a repro is not published. If I cannot hand you the command that shows the thing, the thing does not go on this page. That rule is enforced by a test, not by intention: npm test in this repo fails the build when an entry in the bug log has no repro line under it.

npm test

What is here

• The network sweep: every live host, crawled on a schedule.
• The bug log: what broke, the command that shows it, and whether the fix is confirmed.
• The privacy audit: what our own apps do with your data.
• The flake budget: why a check has to fail twice to count.
• On repro lines: what that rule costs to hold.

The state of the sweep, as of 2026-09-21

No crawl has been published yet. The crawler is written and tested; it has not been run against production from an environment with outbound access to those hosts. The report page says so rather than showing a number I cannot stand behind.

I would rather publish an empty table than a fabricated one. That is most of what this desk is.

The tools behind it

Two of the apps this desk audits with are live and you can use them:

• Diffy: paste two things, see exactly what changed. Nothing leaves the browser.
• CrumbTrail: scans any public page for trackers, analytics scripts, ad pixels, and cookies.

Sweep and HAR are built but not yet live. When they are, they will be linked here and not before.

Who builds this

The AI provider is Anthropic, through Claude Code, used to build, audit, and draft across every first-party repo. (V-5, source: AGENTS.md and CLAUDE.md in calabrodesign, verified 2026-09-21)

That is stated on the front page on purpose. A desk that audits its own work should say who did the work.