@luciezhao Thank you, this is looking great! I added some comments.
question: This customer-facing go/no-go makes sense, but should we add explicitly that if they choose no-go, that any data they created during this stage will be lost after the switchback?
question: Should we move this to Stage 1?
nitpick (non-blocking):
- Rollback is **not a product requirement**; we rely on feature parity + fix-forward.question: Should we add for Geo-based migrations? We know that eventually we want to be able to migrate easily and quickly from any cell to any cell, but I don't think we will be doing that for Geo-based migrations. I guess Protocells implies "initially", but we might be using Geo-based migrations after we drop Proto-.
We will **not** build a general-purpose reverse migration capability for Geo-based migrations. Instead:issue: I think we should not define rollback this specifically, since people already say rollback as a generic word which includes what is being called switchback. In fact the following heading Cell-level rollback seems to use the word rollback in the generic sense.
We could replace this with Reverse migration maybe?
issue: These two lines feel particularly AI? They feel redundant with the content before it, and they feel like decisions more than goals.
And maybe the Bounded scope heading, though its description is a reasonable goal. Though, are these goals for the ADR, or goals that the ADR is trying to resolve?
Should we remove this section?
Add consistent CI/linting infrastructure matching other skill repos.
.gitlab-ci.yml: runs markdownlint-cli2 on every push.markdownlint-cli2.jsonc: rules tuned for AI prompt documentation (long lines, duplicate headings, bare URLs all allowed)lefthook.yml: pre-commit lints staged .md files; pre-push lints all .md files.editorconfig: LF, UTF-8, 2-space indent, trim trailing whitespaceMichael Kozono (6ba25b08) at 14 Mar 08:53
refactor: rename mr-review-many-agents to deep-review-skill
Michael Kozono (5391f60b) at 14 Mar 08:39
feat: add --dry-run flag to install.sh
Michael Kozono (8f8947fd) at 14 Mar 08:32
fix: resolve shellcheck warnings and apply shfmt formatting
... and 4 more commits