Content
77%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 skill body is exceptionally actionable with a rigorously validated 11-step workflow — concrete commands, exact tool parameters, and full worked examples throughout. Its weaknesses are repetition of the same spawn/worktree constraints in multiple places and a monolithic single-file layout where the prompt template and example belong in separate reference files.
Suggestions
State the "spawn_task forks a fresh throwaway worktree off the default branch; the session starts cold" constraint once in the intro and reference it from steps 9-11 and the Important section instead of restating it each time.
Move the full spawn-prompt template and the MES-4282 worked example into references/spawn-prompt-template.md and reference it one level deep, keeping only the field-by-field requirements inline in step 9.
Trim the rationale commentary in step 4b and the Important section to the operative rules (never git add -A, never push, no-op vs. failure handling), which are already repeated in both places.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient — the prose conveys genuinely non-obvious design rationale (worktree object-DB sharing, husky bypass reasoning) — but the core "spawn_task forks a fresh throwaway worktree off the default branch / starts cold in the wrong worktree" fact is restated in the intro, step 9's template, the worked example, step 11's output, and the Important bullets. Fits "Mostly efficient but includes some unnecessary explanation or could be tightened"; not 2 because the length is largely justified, not 4 because the near-verbatim WARNING duplication is real padding. | 3 / 5 |
Actionability | Fully executable throughout: exact bash snippets (git rev-parse --git-common-dir, HUSKY=0 git commit --no-verify), exact tool invocations with parameter shapes (mark_chapter {title, summary}; spawn_task title/tldr/prompt), deterministic title formats with worked examples ([MES-4282] [MR#29] Implementation), exact output text, and a complete MES-4282 sample prompt. Copy-paste ready and covers the common cases. | 5 / 5 |
Workflow Clarity | Eleven clearly sequenced steps with explicit validation at every fragile point: STOP if on the default branch (step 1), guarded no-op vs. loud WARNING commit-failure distinction (step 4b), re-entry append semantics, a one-click user gate with --unattended/--no-handoff escape hatches (step 10), and orphan-worktree pruning guidance. Matches the anchor with explicit validation steps and feedback loops for error recovery. | 5 / 5 |
Progressive Disclosure | The body has section structure (## Steps, ## Important) but no bundle files exist and everything is inline across ~290 lines; the full prompt template and the ~40-line worked example are content that clearly belongs in a one-level-deep references/ file. Fits "Some structure but could be better organized; content that should be separate is inline"; not 4 given zero reference files at this length. | 3 / 5 |
Total | 16 / 20 Passed |