Content
57%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 body is a maximally lean discovery stub that delegates all substance to an external SKILL.md via a repo-relative path escaping the skill bundle. Brevity is excellent, but actionability, workflow clarity, and progressive disclosure all suffer because the actual guidance is entirely absent from this file and the referenced target is not part of the bundle and does not resolve from this location.
Suggestions
Inline the essential evidence-tier rules (or at least a quick-start summary of EP selection, registration, and conflict-resolution steps) in this body so the skill remains actionable when the external path does not resolve.
Make the reference resolve within the skill's own bundle - e.g., bundle the canonical SKILL.md as a references/ file next to this stub - instead of the fragile repo-relative '../../../.agents/skills/windows-ml/' path.
Add a fallback instruction for when the linked file is missing (e.g., where the canonical copy lives or what to do instead) so the single workflow step cannot dead-end.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Three sentences with zero padding: 'The skill lives in ... Read that SKILL.md now and follow its evidence-tier rules. This file only exists so Claude Code can discover the skill.' No over-explanation of known concepts; every token earns its place. | 5 / 5 |
Actionability | The only guidance is the directive 'Read that SKILL.md now and follow its evidence-tier rules' - concrete as a link, but the specific steps to execute (evidence-tier rules, EP selection, conflict resolution) are entirely absent from this body, delegated to '../../../.agents/skills/windows-ml/SKILL.md', a path outside the skill bundle that does not resolve from this file's location. Not 1 because the link and directive are specific and unambiguous; not 3 because no domain guidance exists in the body at all. | 2 / 5 |
Workflow Clarity | The single delegation step ('Read that SKILL.md now and follow its evidence-tier rules') is clear and unambiguous, but this is not a simple skill - the description promises multi-step evidence-tier evaluation, EP registration, and upgrade workflows, none of which are sequenced or checkpointed in this file; all sequencing is implicit in an external file. The simple-skill exception does not apply, and validation checkpoints for the real work are missing. | 3 / 5 |
Progressive Disclosure | The body is a single, clearly-signaled, one-level-deep reference, which is structurally sound. However, it points outside the skill's own bundle ('../../../.agents/skills/windows-ml/SKILL.md') with no references/, scripts/, or assets/ directories alongside, and the target does not resolve from this file's location - navigation to the actual content cannot be verified. Not 2 (nothing is inlined and the reference is prominent, not buried); not 4 (the gap is more than minor - all skill content lives outside the bundle). | 3 / 5 |
Total | 13 / 20 Passed |