Content
86%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.
A tight, well-structured delegation skill: lean instructions, a copy-ready agent prompt, an explicit preserve-list, and explicit per-step confirmation requirements for destructive actions. The only gaps are the absence of spawn syntax and an explicit error-recovery loop if a cleanup step is blocked.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The 13-line body is lean and assumes Claude's competence: an explicit preserve-list ('CLAUDE.local.md', '.claude/settings.local.json', '.claude/worktrees/', ...) and pointed reporting rules with zero padding or explanation of known concepts. Matches the 'every token earns its place' anchor. | 5 / 5 |
Actionability | Gives the verbatim agent task string ('Run the full post-merge cleanup workflow. Preserve all personal and tool-local files...') and concrete preserve/confirm lists, but no syntax for spawning the custom agent and it leans on the agent's own instructions. Mostly executable with minor gaps — anchor 4, not 5. | 4 / 5 |
Workflow Clarity | Five clearly sequenced steps with explicit validation for the destructive operations: 'confirm the merged PR, uncommitted-change handling, branch deletions, force deletions, remote deletions, and origin/master overwrite separately' and 'do not imply the repository is clean unless the final checks prove it'. This avoids the destructive-operation cap of 3, but there is no fix-and-retry error-recovery loop, so anchor 4 rather than 5. | 4 / 5 |
Progressive Disclosure | A simple, single-purpose skill under 50 lines with no bundle files and no external references needed; the numbered list is well organized. Per the simple-skill guideline, well-organized content alone earns the top score. | 5 / 5 |
Total | 18 / 20 Passed |