Flux Controller Patch Releases

Use this skill for upstream Flux controller patch releases only. Do not use it for flux2, pkg, or other non-controller repos.

Supported controllers:

• helm-controller
• image-automation-controller
• image-reflector-controller
• kustomize-controller
• notification-controller
• source-controller
• source-watcher

Important rules

• Go deep until you block, then switch. Drive one controller all the way

through every step you can do locally — branch, changelog commit, version bump commit, push, open PR — before switching to the next. Only move on when you hit something you cannot progress (CI running, PR awaiting review, tag workflow running). The moment something unblocks (PR merged, CI green, workflow finished), come back to it immediately — do not finish the current controller's local work first if an earlier one is ready to advance.

• Never block the conversation on long-running operations. CI checks,

release workflow runs, tag-triggered workflows, and similar waits must be watched in the background so the user can keep steering and so you can pick up any other controller the moment it unblocks. When a background watch completes, report the result and proceed.

• Always quote PR/issue links as full URLs (e.g.

https://github.com/fluxcd/helm-controller/pull/1465), never the <owner>/<repo>#<number> shorthand — full URLs are clickable from the user's terminal, the shorthand is not.

• Start background watches on every PR immediately after opening it. Kick

off gh pr checks <num> -R fluxcd/<repo> --watch in the background as soon as the PR is created so CI status lands in the conversation the moment it finishes. Do the same for tag-triggered release workflows (gh run watch <id> -R fluxcd/<repo> in the background). Do not wait until "everything is pushed" to start watching — start watching the first PR while you prepare the second. Note that gh pr checks --watch can finish early, reporting success on a set of checks that does not yet include a run it just re-triggered, so restart the watch (and re-read gh pr checks <num>) after any amend or force-push.

• Also start a background approval watch per PR. gh pr checks --watch

only covers CI; it does not fire on maintainer approval. Poll the review state in the background so you are notified the moment it flips to APPROVED + CLEAN:

  while :; do
    state=$(gh pr view <num> -R fluxcd/<repo> --json mergeStateStatus,reviewDecision --jq '.reviewDecision+" "+.mergeStateStatus')
    case "$state" in "APPROVED CLEAN") echo "$state"; break;; esac
    sleep 30
  done

Run this in the background; when it exits, merge the PR and proceed. The session's git user opens these PRs, so they cannot approve them: request a review from a maintainer with gh pr edit <num> -R fluxcd/<repo> --add-reviewer <login>, so the approval watch has a reviewer to fire on.

• Every git commit must use -s (sign-off). Never include Co-Authored-By

lines, your own name, or any AI attribution in commit messages, PR titles, or PR descriptions. This applies to all PRs, including PRs that update this skill file itself.

• Always wait for CI to go green before merging any PR.
• You cannot approve your own PRs. If a PR was opened by the git user driving

the session, ask a maintainer to approve it (or confirm it is already approved) before merging.

• Merge release PRs yourself (controller release PRs, changelog cherry-pick

PRs) as soon as CI is green and a maintainer has approved. No need to ask the user to click merge — act on it immediately so the next step (tag push, etc.) unblocks. This applies only to PRs opened with the user's account during this session.

• Review feedback on release PRs is applied by amending, not by adding

new commits. A release PR must stay at exactly two commits (Add changelog entry for vX.Y.Z and Release vX.Y.Z). When the fix belongs in the changelog, amend the changelog commit; when it belongs in the release bump, amend that one. Use git reset --soft HEAD~2 + re-commit, or an interactive rebase, then git push --force-with-lease.

• **After applying a review fix, reply Fixed, thanks! on the thread and

resolve it.** Reply via gh api repos/<owner>/<repo>/pulls/<n>/comments/<cid>/replies -f body='Fixed, thanks!' and resolve via the GraphQL resolveReviewThread mutation. Find thread IDs with gh api graphql -f query='{ repository(owner:"<o>",name:"<r>") { pullRequest(number:<n>) { reviewThreads(first:50) { nodes { id isResolved comments(first:1){nodes{databaseId body}} } } } } }'.

• Do not watch CI on the skill-update PR continuously — it only needs to

merge at the very end of the procedure, so check CI right before merging rather than keeping a watch open throughout the session.

• Tags must be annotated and signed (git tag -s -m ...). Never create

release tags through the GitHub API — that produces lightweight tags which break git tag -v verification.

• Strictly follow the git commands documented in the release flow below. Do

not invent substitutions or skip steps — each step has a reason.

• **Do not declare a controller "done" until every step in the Release Flow

below has been executed for it*, including the final changelog cherry-pick PR back to `main` (step 11) and that PR being merged. Merging the release PR and tagging is not* the last step, and neither is opening the cherry-pick PR. Before reporting completion, walk through each controller against the numbered steps and confirm each one ran.

• Never halt while there is work you can do. The session is over when

every controller has been through all eleven steps and no PR opened during it is still waiting to be merged. Do not stop to report progress and wait for a prompt when nothing is blocking you — reporting is not a step, and "the PRs are open" is not a finish line. The only legitimate pauses are external: CI still running, a PR awaiting a maintainer's approval, a release workflow in flight. When one of those clears, act on it immediately instead of asking whether to continue.

• PRs opened by this procedure use the release commit subject (Release vX.Y.Z)

as the PR title and an empty body — never the changelog commit subject.

Preconditions

• Read the upstream procedure at website/content/en/flux/releases/procedure.md,

section Controllers: patch releases.

• Treat git and gh commands as confirmation points if the user wants that.
• Use date to get the release date for changelog entries.
• Fetch before reasoning about release branches or merged PRs. Do not trust stale

local origin/* refs.

Determining which controllers need a patch release

Do this first, for every supported controller, before touching any of them. Compare the tip of each controller's latest release series branch against the most recent tag on that series. If the branch is ahead of its tag, it needs a patch release and the next version is the next patch increment of that tag.

for repo in helm-controller image-automation-controller image-reflector-controller \
            kustomize-controller notification-controller source-controller source-watcher; do
  cd ~/Documents/github.com/fluxcd/$repo
  git fetch --all --tags --prune
  br=$(git for-each-ref --format='%(refname:lstrip=3)' 'refs/remotes/origin/release/v*.x' | sort -V | tail -1)
  series=${br#release/}; series=${series%.x}
  tag=$(git tag -l "${series}.*" --sort=-v:refname | grep -v '\-' | head -1)
  ahead=$(git rev-list --count "${tag}..origin/${br}")
  echo "$repo  $br  latest=$tag  ahead=$ahead  next=${tag%.*}.$(( ${tag##*.} + 1 ))"
done

Read the output as follows:

• ahead=0 — nothing to release for that controller; skip it.
• ahead>0 — the controller needs a patch release at the printed next version.
• Only the latest series (highest vN.M) is an automatic patch candidate.

Older supported series are patched only for critical fixes, and picking one is a deliberate decision — never infer it from the loop alone.

• A series can be ahead>0 because a release PR is already open for it. Check

gh pr list -R fluxcd/<repo> --state open --base release/vX.Y.x before starting a new one, and resume that PR instead of opening a duplicate.

• Confirm the latest tag against GitHub with

gh api repos/fluxcd/<repo>/git/matching-refs/tags/v --jq '.[].ref' — local tags and refs go stale when a fetch fails, which silently hides or invents unreleased commits.

Release Flow

For each controller identified above as needing a patch release:

1. Refresh local state.

• git fetch --all --tags --prune
• git switch release/vX.Y.x
• git pull origin release/vX.Y.x

1. Create the release preparation branch exactly from the release series branch.

• git switch -c release-vX.Y.Z release/vX.Y.x

1. Draft the new CHANGELOG.md entry.

• Use the existing changelog structure in that repo.
• Use the current date from date.
• Build the entry from the commits merged into the release series branch since

the previous tag.

• Prefer PR titles over commit subjects for bullets.
• Check the PRs that introduced the changes; do not infer titles from local

commit messages.

• For dependency-update PRs, inspect the PR body, any referenced upstream

PRs, and the go.mod diff for notable content (see "Dependency update PRs" below) rather than settling for a generic bump line.

• In the intro paragraph, summarize the concrete bug fixes shipped in the

patch release.

1. Commit the changelog entry.

• git add CHANGELOG.md
• git commit -s -m "Add changelog entry for vX.Y.Z"

1. Apply the release version bump exactly as documented.

• Update the controller self-API version in the root go.mod.
• Update config/manager/kustomization.yaml newTag to vX.Y.Z.
• Commit with:
• git add go.mod config/manager/kustomization.yaml
• git commit -s -m "Release vX.Y.Z"

1. Push the release preparation branch.

• git push origin release-vX.Y.Z

1. Open and merge the release PR into the release series branch.

• Base: release/vX.Y.x
• Head: release-vX.Y.Z
• Merge with a merge commit (gh pr merge <num> -R fluxcd/<repo> --merge),

not a squash: the tagged commit on the release series branch must keep both signed commits and the DCO history.

1. Refresh the release series branch after merge.

• git switch release/vX.Y.x
• git pull origin release/vX.Y.x

1. Create and push signed tags from the updated release series branch.

Push the api/ tag first — the release tag depends on it.

• git tag -s -m "api/vX.Y.Z" api/vX.Y.Z
• git push origin api/vX.Y.Z
• git tag -s -m "vX.Y.Z" vX.Y.Z
• git push origin vX.Y.Z

1. Confirm the non-api/ tag triggered the release workflow.

• It runs release.yaml on the vX.Y.Z tag; wait for it with

gh run watch <id> -R fluxcd/<repo> --exit-status, then confirm a non-draft GitHub release was published (gh release view vX.Y.Z).

1. Cherry-pick only the changelog commit back to main.

• git switch main
• git pull origin main
• git switch -c pick-changelog-vX.Y.Z main
• git cherry-pick -x <Add changelog entry commit>
• git push origin pick-changelog-vX.Y.Z
• Open PR from pick-changelog-vX.Y.Z to main
• Do this one controller at a time — never chain several controllers'

cherry-picks into a single shell invocation. If such a batch is interrupted partway, the branches it already created make a later git switch -c fail, which silently leaves you on main, so the next git cherry-pick lands the changelog commit on local main instead of the pick branch. Before cherry-picking, confirm you are on pick-changelog-vX.Y.Z; if the branch already exists, switch to it (or delete and recreate it) rather than letting switch -c fail.

How To Build The Changelog Entry

For a patch release, gather:

• the latest release tag on the release line
• the merged commits on origin/release/vX.Y.x since that tag
• the PRs corresponding to those merges

Write the new section at the top of CHANGELOG.md:

• ## X.Y.Z
• **Release date:** YYYY-MM-DD
• short intro paragraph describing the actual bug fixes in user-facing language
• Fixes: when there are bug-fix items
• Improvements: for dependency updates, docs, feature gates, or cleanup

Rules:

• Use PR titles, not raw commit messages
• Group multiple dependency bump PRs naturally when the repo history already does that
• Verify the title against GitHub when the local merge commit is vague

Dependency update PRs

Do not reduce a dependency bump to a generic "Update fluxcd/pkg dependencies" line and move on — the user-facing substance is usually hidden inside the bump. Weigh Flux's own repos and third-party dependencies differently:

• *The `fluxcd/` modules are what matter.** These are our own libraries, so a

bump is how a fix we made reaches the controller. Trace every module in the go.mod diff back to its commits, e.g. in a local clone of fluxcd/pkg: git log --oneline --no-merges <mod>/<old>..<mod>/<new> -- <mod>, filtering out Prepare for release and the shared Upgrade k8s to ... commits. Read the PR behind each remaining commit and describe the fix in the controller's own terms. Doing this across all the modules of one release usually leaves a handful of real changes, which is a short enough list to reason about per controller.

• Third-party bumps are noise unless they fix something our users hit. Do

not recount an upstream project's release notes, and do not list upstream CVE/GHSA advisories that merely rode along with a bump. Mention an upstream change only when it resolves an issue reported by Flux users, or when the upgrade itself is the point (e.g. syncing the Helm version with the one helm-controller ships) — and then say why it matters to Flux, not what changed upstream.

• *State the Kubernetes bump when `k8s.io/` actually moves**, e.g. "which

bring Kubernetes to 1.36.4". Read the version off the go.mod diff for that specific bump instead of assuming one is there: a fluxcd/pkg update may carry a Kubernetes bump, a module fix, both, or neither, and the same release window can move k8s.io/* for one controller and not for another.

• Check whether the changed code is reachable from the controller, and be

as careful about wrongly leaving it out as about wrongly putting it in. A fix can land in a function the controller never calls, in which case it does not belong in that changelog; or in one the controller already calls, in which case the behavior change ships without any code change in the controller and must be in the changelog. grep the controller for the changed symbol and for the type that owns it. Example: runtime/client rejecting file references in kubeconfigs reached kustomize-controller purely through Impersonator, which kustomize-controller uses and the other controllers do not — so it changed how .spec.kubeConfig Secrets are accepted with no kustomize-controller commit behind it.

• When several controllers pull the same bump, use matching wording across

their changelogs so readers can correlate them.

• Only mention a dependency change if it is relevant to what that controller

actually does. A bumped module often ships capabilities the controller never exercises, and listing them implies a capability it does not have. For example, GCP sovereign cloud artifact registry support (shipped via a pkg/auth bump) is worth noting for source-controller and image-reflector-controller, which pull from registries, but not for image-automation-controller, which only talks to Git.

Critical Checks

• Always fetch before comparing tag..origin/release/....
• Always pull the release series branch before creating release-vX.Y.Z.
• Always pull the release series branch again after merging the release PR and

before tagging.

• Always inspect the actual root go.mod; do not assume the self-API path form.

source-watcher uses github.com/fluxcd/source-watcher/api/v2, so it still needs the same self-API release bump pattern.

• Do not silently special-case a controller. If the documented step seems not to

apply, inspect the file and confirm before proceeding.

• Tag from the release series branch merge commit, not from the release prep branch.
• Cherry-pick only the changelog commit back to main, not the release version bump.
• When amending a release PR, confirm the version bump survived. After

git reset --soft HEAD~2, do not git checkout go.mod or config/manager/kustomization.yaml — that restores them from the index and silently reverts the bump, leaving a release PR that only touches the changelog. Rebuild both commits, then check with git diff origin/release/vX.Y.x --stat that CHANGELOG.md, go.mod and config/manager/kustomization.yaml are all still modified.

Updating this skill

• Improvements to this skill should land as a single-commit PR on a dedicated

branch. When accumulating more changes during a release session, amend and force-push rather than adding new commits.

• Keep the skill-update PR open during the release session and merge it

last, after all controller patch releases are done. Session learnings tend to surface throughout the flow; amend them in as they come up.

• Do not leak session-specific state, downstream/enterprise distribution

details, or AI attribution into the skill file.

Useful Local Queries

• Release branches:
• git branch -r --list 'origin/release/v*.x' | sort -V
• Latest tags:
• git tag -l 'v*' | sort -V | tail
• Commits since previous release on the release branch:
• git log --oneline <prev-tag>..origin/release/vX.Y.x
• PR metadata for changelog bullets:
• gh pr view <number> -R fluxcd/<repo> --json number,title,url,baseRefName