Start with the test, because your list will not be identical to anyone else's and the test travels better than the list.
A design ritual belongs on the kill list when three things are true at once. It exists to produce a receipt rather than a decision. The receipt was proof that judgment happened, back when producing it was expensive. And producing it is no longer expensive.
All three. Plenty of design rituals pass that test fine. Calibration passes. A failure inventory passes easily.
Nine fail, and each one gets a replacement in the same breath.
Pixel-perfect handoff dies, replaced by a constraint set plus a range spec delivered before generation instead of after design. Engineering gets rules and boundaries rather than coordinates, which is faster to write and survives the first product change that would have invalidated every measurement in the file.
Design review as approval theater dies, replaced by a thirty-minute rubric review in the same calendar slot. Scores in before, disagreements during, one decision at the end.
The component library as monument dies, replaced by a constraint layer and a much smaller set of primitives. When surfaces get assembled at runtime, the value sits in the rules that govern assembly rather than in the count of pre-built components.
Personas as decoration die, replaced by an evidence register with sources, strength, and expiry dates. When somebody asks who this is for, the answer cites a row instead of a character.
The double diamond used as a quarterly calendar dies, replaced by a weekly rhythm of one question, one register row, one decision. Fidelity ladders die, replaced by going straight to whatever fidelity answers the question in front of you. The design QA pass dies, replaced by automated constraint checks plus the rubric review, because design inspecting output at the end of a cycle is a QA function with better typography. The two-day kickoff workshop as default dies, replaced by ninety minutes on a failure inventory with three people, which produces the constraints and the eval cases and gives you the rest of the two days back. And feedback rounds with no decision rule die, replaced by a named owner, named constraints, and a stated question: here is the decision, here are the constraints that produced it, tell me which constraint is wrong.
The list is the easy part. What makes a kill stick is the protocol, and skipping its second step is why most kills fail.
Name what the ritual was protecting. Every surviving ritual protects something real. Handoff protected the specificity of the built result. Review protected quality. Say it out loud, or the people who value it will assume you did not notice, and then the argument is about whether you understand the work.
Ship the replacement first, and run both for two weeks. Overlap is the price of a kill that sticks.
Kill the meeting, not just the artifact. A ritual with a standing calendar slot regenerates its artifact, and you will be explaining six weeks later why the deprecated template is somehow back.
Say when you would bring it back. Name the condition out loud on the day you announce the kill. That sentence turns an ideology into an experiment, and it is what gets the skeptics to agree.
One thing to do this week: pick the ritual on your team that everybody privately thinks is theater, write down what it protects and what replaces it, then run both for two weeks and announce the bring-back condition on day one. The full nine with a blank scorecard are in The Kill List, and the long argument is in Design's Kill List.