Repository navigation
Release-prep skill based on princeton-cdh software release checklist - #7 - #1
Conversation
Created with the help of claude-sonnet-4-6
Assisted-by: Claude:claude-sonnet-5
tanhaow
left a comment
There was a problem hiding this comment.
I thoroughly enjoyed the experience of preparing a release using this agentic skill. Based on my observations and experiences, I have a few suggestions:
- Add the
rcpostfix to the version number. - Open the PR as a draft PR instead of a regular PR.
- Ensure that the changelog references previous entries to maintain consistency in language and style.
- Keep the “closes #xxx” section to automatically connect the PR to the issue on Zenhub.
Thank you once again for spending time writing this skill up. It’s very convenient and helpful!
|
Just found that we might want to include one more task in the "Needs your judgment" checklist section that is to decide which PRs by |
laurejt
left a comment
There was a problem hiding this comment.
My experiences using the skill with a local model (gemma4) for viapy (minor changes since the last release):
- Overall: Not sure how well this skill will work for local models, in the end all this did was bump the version and make/push a release branch
- It failed to run the unit tests because it didn't figure out the need to use
uv run. Seems like it might be worth an elaboration on environment assumptions, since the skill suggestspytestis installed in the active environment instead of say viauv run - Only change in release branch was to bump the version
- No changelog edits. Partial failure due to not having authenticated
gh, but clearly did not add anything based on the commit log. - Got an odd local settings flagged message: "The detected sample file path was outside the current repository's context, preventing a comparison against local settings." I'm guessing this has something to do with viapy's django unique/optional django setup (only sample setting file is for unit tests)
- It did check the uv.lock, so that's something
- Says a PR was created, when it hasn't been (but the release branch was pushed)... seems like the TUI output might be the text intended for the PR message
- The commit make no mention of automation, which seems important since these commits will end up in the main commit history
Question: What is (or should be) the expected behavior of the skill if a pre-commit hook fails? It seems plausible that this might occur when drafting the changelog.
Some suggestions:
- Add step for reviewing code documentation? It's in the checklist but perhaps it's out-of-scope for the skill
- Add step for reviewing the commit log? It's in the checklist and it seems plausible that the skill could flag commits that should be combined or whose logs need some editing.
- It seems like there should be some accompanying human-centered documentation that provides information about the skill, particularly the general expectation for a config file.
Co-authored-by: Hao Tan <tanhaosmail@gmail.com>
Co-authored-by: Hao Tan <tanhaosmail@gmail.com>
Assisted-by: claude.ai:claude-sonnet-5.5
Assisted-by: OpenCode:claude-opus-5.5
Assisted-by: OpenCode:claude-opus-5.5
Changes in this PR
Notes
Reviewer Checklist