rebase-clean

name	rebase-clean
description	Performs a strict clean rebase of a feature branch onto main with minimal conflict resolution and full validation. Use when the user asks to rebase carefully, avoid extra branches, avoid exploratory edits, and run go test and go vet until green.
disable-model-invocation	true

Clean Rebase Workflow

Use this workflow exactly for rebases in this repository.

Rules

• Do not create extra temporary branches unless the user explicitly asks.
• Do not make unrelated edits.
• Do not stop after analysis; finish rebase and validation.
• Keep output brief and operational.
• Never assume target branch names; detect current branch first.
• Use go test ./... -count=1 from the repo root (this repo has no make test wrapper).

Step 0: Detect Branch Context

1. Run git branch --show-current.
2. Treat detected current branch as the candidate target branch.
3. Ask user to confirm explicitly before proceeding:
   • I detected '<current-branch>' as the current branch. Should I rebase this branch?
4. If user specified a target branch, verify it matches current branch; if not, stop and ask.
5. Do not run reset/rebase commands until this confirmation is received.
6. Before any reset/rebase command, restate:
   • current branch
   • target branch
   • approved baseline ref and confirm when there is any mismatch or ambiguity.

Step 1: Restore Branch Baseline (Rerun Only)

Use this step only when a prior rebase attempt failed or introduced unwanted edits. For a first rebase attempt, skip this step and go directly to Step 2.

1. Checkout target branch.
2. Hard reset to the known backup/base commit the user approved.
3. Fetch upstream.

Command pattern:

git checkout <branch>
git reset --hard <backup-or-approved-base>
git fetch upstream

Step 2: Rebase Onto Main

Run:

git rebase upstream/main

If conflicts occur, resolve only the conflicted files. Do not add extra refactors.

Step 3: Conflict Resolution Policy

For each conflicted file:

1. Start from one side (ours or theirs) as a temporary base.
2. Apply only minimal compatibility changes required to compile and run tests on the new base branch.
3. Keep behavior from the feature commit unless it is incompatible with upstream removals/renames.
4. Resolve matching tests together with production code changes.
5. Avoid opportunistic refactors during conflict resolution.
6. Go module conflicts: If go.mod or go.sum conflict, prefer theirs (upstream) and run go mod tidy after.

Then:

git add <resolved files>
GIT_EDITOR=true git rebase --continue

Repeat until rebase completes.

Step 4: Verify No Code Loss

Run these checks against the approved pre-rebase baseline (backup/base commit):

git range-diff upstream/main...<approved-base> upstream/main...HEAD
git diff --name-status <approved-base>..HEAD
git merge-base --is-ancestor upstream/main HEAD
git log --left-right --cherry-pick --oneline upstream/main...HEAD

Acceptance criteria:

1. Branch intent is preserved:
   • Commit intent maps cleanly in range-diff (rewritten SHAs are fine).
2. Main is fully incorporated:
   • git merge-base --is-ancestor upstream/main HEAD succeeds (exit code 0).
3. No unexpected file removals/renames beyond what upstream already changed.
4. Any conflict-touched files have expected compatibility-only deltas.
5. No unexplained drift from the approved baseline.

If any check fails, treat as code-loss risk and restart from Step 1.

Step 5: Regenerate CRDs (If API Changed)

If any files in api/v1alpha1/ were touched during rebase, CRDs are generated in lightspeed-operator (see repo CLAUDE.md):

cd ../lightspeed-operator
make manifests-agentic

Commit the regenerated CRD YAML in the appropriate repo; amend the feature commit only if that matches the user’s workflow.

Step 6: Full Validation Pipeline

Run:

go test ./... -count=1 && go vet ./...

Optionally run the API kube linter if api/ changed (see validate-and-fix skill).

If validation or code-loss checks fail at any point:

1. Treat it as an incorrect rebase result.
2. Abort/restore to the approved baseline.
3. Re-run the rebase/conflict-resolution flow from Step 1.
4. Re-run Step 4 and this step.
5. Repeat full loop until all checks pass.

Step 7: Optional cluster / E2E validation

Only if the user requests and has a cluster setup documented for this repo (no default make test-e2e target here).

Step 8: Final Report

Report only:

• Rebase completed (yes/no)
• Current branch and ahead/behind state (git status -sb)
• go test ./... -count=1 result
• go vet result
• Cluster/E2E result (if run)
• CRDs regenerated via lightspeed-operator (yes/no/not needed)

Do not include unrelated diagnostics.