Content
92%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is lean, actionable, and clearly sequenced with explicit verification checkpoints. Its one structural weakness is progressive disclosure: the SKILL.md points to references/ files that are not present in the bundle, leaving the reader unable to actually load the persona or reviewing details it depends on.
Suggestions
Ship the referenced bundle files (references/reviewer-selection.md, references/reviewing.md, and references/reviewers/<persona>.md) so the documented load steps can actually be executed.
If the persona files are intentionally absent, inline the minimum persona-selection criteria needed to spawn reviewers instead of pointing at a missing file.
Verify each in-body link resolves to a real file under the skill package; dangling references undermine both navigation and actionability of the load step.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and operational — it does not explain concepts Claude already knows (no "what is a PR/diff"), and every section (Contract, Before You Start, Workflow, Output) carries concrete guidance rather than padding, matching the top anchor for assuming Claude's competence. | 3 / 3 |
Actionability | Provides copy-paste-ready material: concrete commands ("git diff --stat <base>...HEAD", "pnpm test path/to/spec"), a fill-in subagent prompt template, and a concrete output example, satisfying the top anchor for fully executable guidance. | 3 / 3 |
Workflow Clarity | Clear sequenced workflow (5 prep steps, 4 numbered phases) with explicit verification emphasis — "Run the smallest runtime check that changes the verdict", "Treat stale CI... as unverified until refreshed", and "Block when missing context or proof prevents an honest verdict" — giving validation checkpoints and an unverified-to-verdict feedback loop. | 3 / 3 |
Progressive Disclosure | References are well-signaled and one-level-deep (references/reviewer-selection.md, references/reviewing.md, references/reviewers/<persona>.md), but the referenced bundle files and directories do not exist, so the split is aspirational rather than real; scoring against the actual (empty) bundle structure pulls this below the top anchor. | 2 / 3 |
Total | 11 / 12 Passed |