Content
92%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is an excellent example of skill content: fully executable commands, a decision procedure with look-first checks, and unusually thorough validation before the risky rotation step. Its only flaw is a small duplicated explanation of squash-merge behavior across two sections.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes git competence throughout, but the squash-merge explanation ("squash merges leave the source commits off `main`'s history") is stated twice — once in Create and again in After a merge — a minor duplication that could be trimmed. | 4 / 5 |
Actionability | Every command is real and executable (the name-generation bash script is copy-paste ready), and runtime placeholders like <mergeCommit.oid> and <ship_merge_head_oid> each come with a documented source command (gh pr view --json mergeCommit / headRefOid). | 5 / 5 |
Workflow Clarity | Create opens with look-first checks and a three-case decision procedure, and After a merge specifies explicit validation (mergeCommit ancestry, clean-tree check with path excludes, unpushed-commit check, headRefOid match, cat-file/merge-base proofs) plus error-recovery feedback ("wait a few seconds and fetch again", "keep the branch"). | 5 / 5 |
Progressive Disclosure | This is a compact single-purpose skill with no bundle files; its three sections (Name, Create, After a merge) are well organized, nothing that belongs in a separate file is inlined, and the one cross-skill pointer (babysit-pr) is clearly signaled. | 5 / 5 |
Total | 19 / 20 Passed |