fix(#161): de-hardcode master in CONTRIBUTING and CI workflows#162
Conversation
Addresses three P1 codex findings posted post-merge on the #156 batch: - CONTRIBUTING.md said "Branch from master" — regressed the <default-branch> convention SKILL.md adopted in PR #155. - version-bump-check.yml and release.yml hardcoded branches: [master], which would silently no-op if the repo were ever renamed to main. Fix uses <default-branch> placeholder in docs and lists both [master, main] in workflow filters (Actions does not template the branches filter, so a literal list is the correct shape). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
|
To use Codex here, create an environment for this repo. |
|
@codex review |
|
Codex Review: Didn't find any major issues. More of your lovely PRs please. ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
[shiplog/review] #162: Cross-model review — approve via codex 👍 + repo owner authorizationThis time we waited for codex before merging. Codex completed its review on PR #162 and returned a clean signal:
Per Authored-by: claude/opus-4.7 (claude-code) |
Summary
Fixes the two cohesion regressions codex flagged on PRs #158/#159/#160 after they had already been squash-merged: hardcoded
masterreferences in CONTRIBUTING.md instructions and in both new GitHub Actions workflows.Closes #161
Journey Timeline
What we discovered (post-merge)
After the four-PR #156 batch was merged in rapid succession, the codex bot completed its review and posted P1 line-comments on the three closed PRs:
master" hardcodes a branch name; the project moved to<default-branch>language in PR fix(#154): use <default-branch> in SKILL.md Step 2 walkthrough #155 four days earlier.branches: [master]filter would silently no-op if the repo were ever renamed tomain.branches: [master]brittleness on the release workflow.Codex's stated premise (that shiplog's default branch is
main) is wrong —gh repo viewconfirms the default ismaster, so today the workflows do fire as intended. But the broader concern is correct: the project's own<default-branch>convention from PR #155 was just split again, and a future rename would silently break CI.Decisions
branches: [master, main]literal listbranches:filter, so two-branch literal is the simplest correct shape. Cheap belt-and-suspenders that survives a rename.<default-branch>placeholder@codex reviewafter merge to backfill. We are doing the former on this one.Changes
CONTRIBUTING.md:master" → "Branch from your<default-branch>".mastertriggers..." → "Each version bump on the default branch triggers...".--target master→--target <default-branch>..github/workflows/version-bump-check.yml:branches: [master]→branches: [master, main]with an inline comment explaining the rationale..github/workflows/release.yml: same change with the same explanatory comment.No changes under
commands/,skills/, or.claude-plugin/— so the version-bump-check workflow this PR adjusts will pass without aplugin.jsonbump (and the absence of one is intentional and documents that the workflow exempts non-gated paths correctly).Testing / Verification
master, which is what this PR targets) and should pass since no gated paths changed.plugin.jsonversion is bumped.README.mdmentions of "master" in unrelated narrative are out of scope here.)Knowledge for Future Reference
chatgpt-codex-connector[bot](with the[bot]suffix in the API).gh pr view --jq '.author.login'strips the suffix; the raw API does not. When automating bot-finding inspection, query the raw API to be sure.<default-branch>(docs) or list bothmaster, main(CIbranches:filters). PR fix(#154): use <default-branch> in SKILL.md Step 2 walkthrough #155 is the precedent.Authored-by: claude/opus-4.7 (claude-code)
Captain's log - PR timeline by shiplog