Content
73%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 well-structured orchestration skill with an unusually clear workflow: scope resolution, parallel launch rules, filtering, ordering, and a path forward are all explicitly sequenced. The main cost is redundancy — the read-only constraint and reviewer role are restated per sub-agent — and the absence of a concrete prompt template keeps guidance one step short of copy-paste ready.
Suggestions
State the read-only constraint once in Step 2 ("Every sub-agent is read-only: no edits, apply_patch, staging, or commits") and delete the two-sentence repetition from each of the four sub-agent sections; likewise state "Recommended sub-agent role: `reviewer`" once.
Add a short literal example of the sub-agent prompt and one example normalized finding (file/line, category, severity, why, fix, confidence) so the launch and output steps are copy-paste ready.
Move the four sub-agent role checklists into references/ (e.g. references/reviewer-roles.md) and summarize each role in one line in SKILL.md, keeping the always-loaded overview lean.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient and assumes Claude's competence (no explanations of git or review concepts), but the constraint "This sub-agent is read-only. It must not edit files, apply patches, or make any other workspace changes" is repeated verbatim in all four sub-agent sections after the same rule already appears in the Step 2 bullet list, and "Recommended sub-agent role: `reviewer`" is repeated four times. This systematic redundancy (~10 lines) goes beyond the 4 anchor's 'minor instances that could be trimmed' and fits the 3 anchor's 'could be tightened'. | 3 / 5 |
Actionability | Concrete, executable guidance throughout: exact diff commands ("git diff", "git diff --cached"), an explicit per-sub-agent prompt checklist ("file and line or symbol, issue, why it matters, recommended follow-up, and confidence"), a normalized finding schema with severity/confidence enums, and named output groupings ("fix now", "fix soon", "optional follow-up"). It falls short of 5 because no literal sub-agent prompt template or example rendered finding is provided — the launcher must assemble prompts from the requirement lists. | 4 / 5 |
Workflow Clarity | Five clearly sequenced steps with explicit decision checkpoints: a scope-precedence order with "If there is no clear review scope, stop and say so briefly", an intent-inference fallback ("infer it from the diff and say that the inference may be incomplete"), aggressive filtering rules, and an escape hatch ("If there are no material issues, say that directly"). The skill is read-only, so destructive/batch validation caps do not apply, and the checklist-plus-decision-gate structure matches the 5 anchor. | 5 / 5 |
Progressive Disclosure | No bundle files exist, and the single-file body is well-sectioned (When to Use, Steps 1-5, Limitations) with no buried or nested references. This fits the 4 anchor's good-structure placement; it is not 5 because the file is ~185 lines and the four sub-agent role checklists, which are only needed at launch time, could arguably live in reference files to slim the always-loaded overview. | 4 / 5 |
Total | 16 / 20 Passed |