Branch Merge Assistant

The skill explicitly forbids unapproved force-pushes, skipped hooks, and rewriting or dropping commits, and asks the user when conflict resolution is uncertain. However, it directly commits uncommitted changes and runs a merge in the main worktree without general pre-execution confirmation, impact disclosure, or a rollback plan after success, so points are deducted. MIT licensing, Microsoft attribution, and repository provenance are traceable governance evidence only and do not resolve the authorization gaps.

The workflow covers status checking, merging, conflict listing, staging, committing, and post-merge validation, with some useful conflict feedback. However, it depends on an unavailable /commit skill, assumes a context-provided branch and worktree setup, lacks handling for general command failures, and provides no skill-specific test evidence. Because this is a static review without execution, the score is kept conservative.

The name, trigger description, and scenario are clear: merge a session topic branch back into its base branch. Boundary evidence is limited because the workflow assumes a supplied context block and a particular current/main worktree arrangement, without defining non-fit cases, permission prerequisites, or precise trigger exclusions. The function is local Git work and does not entirely depend on overseas services; Chinese users can use the commands, but the documentation is not localized.

The document is concise and organized into purpose, safeguards, workflow, conflict handling, and validation, with several explicit restrictions. Repository materials establish MIT licensing, Microsoft ownership, version information, and source location. The skill itself lacks installation or dependency notes, examples, FAQs, known limitations, changelog/versioning, and a clear maintenance or update path, so points are deducted.

The core objective and commands can support a normal merge, and the final checks validate a clean main worktree and ancestor relationship. However, automatic committing, dependence on the separate /commit skill, incomplete abnormal-case handling, and the absence of an expected result report reduce direct usability. Static reading cannot verify execution, so the score remains within the conservative effectiveness ceiling.

The commands, branch relationship, and validation conditions are auditable, and repository files corroborate basic licensing and governance claims. There is no skill-specific test suite, CI coverage, or third-party execution evidence; the supplied CI workflows do not cover the merge skill's key path. Static evidence therefore supports only a limited score.