I haven't read this in detail yet, but this post does a two things I like that I'd like to highlight:

1) Using a google doc as a WIP that people can comment on to work out a practical problem you're in the middle of thinking about (which I think is a good practice)

2) Providing context for a link post so that it's easier to start a discussion.

I think there is a big question of what to do when we have something closed to aligned AI. Who attempts to take it towards super intelligence? How will we make sure it is actually beneficial to humanity not just the sub group of people that are building it. If the builders of it cannot convince the rest of the world of this there will be multiple attempts to build it, which might be bad under some scenarios.

Things have been very quiet on the coherent extrapolated volition front.

Your plan currently only addresses ex-risk from AGI. However, there are several other problems that should be considered if your goal is to prevent global catastrophe. I have recently been reading 80000 Hours, and they have the following list of causes that may need to be included in your plan: https://80000hours.org/articles/cause-selection/

In general, I think that it's difficult to survey a wide topic like AI Alignment or Existential Risk, and, with granularity, write out a to-do list for solving them. I believe that people who work more intimately with each ex-risk would be better suited to develop the on-the-ground action plan.

It is likely that a variety of ex-risks would be helped by reaching for similar goals, in which case, high-level coordinated action plans developed by groups focused on each ex-risk would be useful to the community. If possible, try to attend events such as EA conferences where groups focusing on each of the possible global catastrophes will be present, and you can try to capture their shared action plans.