📊 TypeScript Coverage Report

Coverage: 39.6%

Greptile Summary

This PR adds a new Claude skill (cherry-pick-sync) that automates cherry-picking commits from main into an active release/* branch. The skill is well-structured and follows the pattern established by the existing .claude/skills/release.md — it identifies the target branch, filters commits, creates a working branch, cherry-picks one commit at a time (handling conflicts), and opens a PR. Overall the workflow is sound, but there are a few correctness issues in the instruction text itself that could cause an AI agent to behave unpredictably:

• Branch naming ambiguity: The working-branch template chore/cherry-pick-main-into-<release>-YYYYMMDD doesn't explicitly state that the / in release/2.2.0 should be replaced with -. The inline example shows the correct sanitized form, but the lack of an explicit rule means Claude could generate a branch name with an embedded slash (e.g., chore/cherry-pick-main-into-release/2.2.0-YYYYMMDD), creating an unintended nested ref.
• Contradictory multi-branch instructions: Step 1 says "Multiple → ask the user" while the Edge Cases section says "Default to most recently committed." These two rules conflict and should be reconciled into a single consistent behavior.
• Missing abort/cleanup path: If a cherry-pick conflict cannot be resolved and the user wants to cancel, there are no instructions for git cherry-pick --abort or branch cleanup, risking leaving the repository in a partially-applied state.

Confidence Score: 3/5

• Safe to merge with minor fixes — the skill works correctly in the common case but has two logic-level contradictions in its instructions that could cause inconsistent AI behavior.
• The change is documentation/instruction-only (a SKILL.md file), so there is no runtime code risk to the Kora service itself. However, the skill contains a branch-naming ambiguity and contradictory multi-branch instructions that would cause an AI agent to behave unpredictably in edge cases. These are correctness issues within the skill's logic, not security or data-integrity risks, but they should be resolved before the skill is relied upon in production workflows.
• .claude/skills/cherry-pick-sync/SKILL.md — branch naming template and multi-branch handling instructions need to be reconciled.

Important Files Changed

Sequence Diagram

sequenceDiagram
    participant U as User
    participant C as Claude (Skill)
    participant G as Git
    participant GH as GitHub CLI

    U->>C: "sync main to release"
    C->>GH: gh pr list --head "release/" (open PRs)
    GH-->>C: release branch name(s)
    alt Multiple branches
        C->>U: Ask which branch to use
        U-->>C: Confirm branch
    end
    C->>G: git fetch origin main + release-branch
    C->>G: git log --oneline --reverse origin/release..origin/main
    G-->>C: Commit list
    C->>C: Filter: auto-skip "chore: release v*", CHANGELOG, version bumps
    C->>U: Show final commit list & confirm
    U-->>C: Approve
    C->>G: git status --porcelain
    G-->>C: Clean / Dirty
    alt Dirty
        C->>U: Ask to commit or stash
    end
    C->>G: git checkout -b chore/cherry-pick-main-into-release-X-YYYYMMDD origin/release-branch
    loop Each commit SHA
        C->>G: git cherry-pick SHA
        alt Conflict
            C->>C: Inspect conflict markers
            alt Trivial (whitespace/import)
                C->>G: Auto-resolve + git add . + git cherry-pick --continue
            else Non-trivial
                C->>U: Ask for resolution
                U-->>C: Resolution
                C->>G: git add . + git cherry-pick --continue
            end
        end
    end
    C->>G: git push -u origin HEAD
    C->>GH: gh pr create --base release-branch
    GH-->>C: PR URL
    C->>U: Return PR URL

_{Last reviewed commit: 7c8925d}