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 tight, instruction-only review skill that maximizes signal per token: dense checklists, a sequenced workflow with genuine validation checkpoints, and clean delegation to sibling skills. The only soft spot is a few conceptual framing passages that are slightly less actionable than the concrete checklists.
Suggestions
In the First-Read Pass, the framing "replay the change as if designing it from scratch" could be paired with one concrete first action (e.g., "list every touched file first") to push actionability to fully executable.
The Ownership And Collapse Check poses rhetorical questions; converting the top two into a concrete ordered test (count callers, then check the five ownership reasons) would make the pass more copy-paste runnable.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Lean and dense throughout — every line is a directive or checklist with no padding and no explanation of concepts Claude already knows; assumes competence at every turn. | 5 / 5 |
Actionability | Provides concrete checklists and templates (abstraction marks, one-caller ownership test, ASCII tree, clean-HEAD reproduction) that are directly usable; minor gaps remain where guidance is framing rather than a discrete step. | 4 / 5 |
Workflow Clarity | The 10-step Review Order is clearly sequenced with explicit validation checkpoints — most notably reproducing failures on clean HEAD and separating pre-existing red from regressions — plus a report-before-edit feedback loop. | 5 / 5 |
Progressive Disclosure | Well-organized into focused sections and delegates detail to one-level-deep sibling skill links (collapse-pass, refactoring, code-audit, typescript, etc.) with no nested references; no bundle files exist so structure is self-contained and cleanly partitioned. | 5 / 5 |
Total | 19 / 20 Passed |