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.
A tightly written, senior-level review playbook with concrete commands, a severity-ranked output template, and a real validation gate. The only improvement would be a worked example finding to make the output format fully self-modeling.
Suggestions
Add one short worked example of a finding rendered in the output format (e.g. a sample [severity: high] line with reachable path and fix) so the template is self-modeling.
Optionally surface the diff-base resolution as an explicit first step (e.g. how to determine <base> when unspecified) to remove the one remaining ambiguity in an otherwise executable workflow.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Lean and efficient throughout; assumes Claude's competence (no explanation of git, callers, or basic review concepts) and every section earns its place, including the load-bearing framing in the opener. | 5 / 5 |
Actionability | Provides copy-paste commands ('git diff <base>...HEAD', 'gh pr diff') and a complete output-format template, but lacks a fully-worked example finding rendered in that template to model the common case end-to-end. | 4 / 5 |
Workflow Clarity | Clear sequence (Scope → What to hunt → Confidence gate → Output → Boundaries) with an explicit validation checkpoint in the confidence gate ('Report a finding only when you can name the reachable path') and a 'Considered, not findings' feedback channel; read-only so the destructive-cap does not apply. | 5 / 5 |
Progressive Disclosure | Single-purpose skill with well-organized section headers and no monolithic wall of text; no external references are needed, so clean in-body structure fully satisfies progressive disclosure. | 5 / 5 |
Total | 19 / 20 Passed |