Content
71%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 content is a well-structured, actionable MR-protocol generation framework with excellent progressive disclosure via explicitly mapped, verified reference modules. Its main weakness is redundancy across the Hard Rules, What-Not-To-Do, and Quality Standard sections that restates earlier guidance, and the absence of an explicit output self-validation step or a worked example deliverable.
Suggestions
Consolidate the redundant rules: fold the 'Hard Rules', 'What This Skill Should Not Do', and 'Quality Standard' sections into the relevant execution steps and Mandatory Output Structure to remove restated guidance and reduce token cost.
Add a final 'Self-validation' step (e.g. Step 9) that checks the generated output against the A–L structure and the reference-module usage gate before delivery, with a fix-and-retry loop for any missing section.
Include one short worked example of a completed output (e.g. a condensed Standard-plan protocol for a sample exposure→outcome pair) so the expected deliverable format is concrete rather than implied.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient operational specification (it does not tutor Claude on what MR is), but the Hard Rules, 'What This Skill Should Not Do', and 'Quality Standard' sections substantially restate rules already given in the 8 steps and Mandatory Output Structure (e.g. IVW default, bidirectional/MVMR cautions, claim boundaries), which is more than minor padding and could be tightened. | 3 / 5 |
Actionability | Concrete and specific guidance throughout — verbatim out-of-scope redirect text, sample input triggers, a fixed A–L output structure, named methods (IVW, weighted median, MR-Egger, MR-PRESSO, Steiger), and explicit per-section requirements — but it lacks a fully worked example of a completed output, leaving a minor gap for a generation skill. | 4 / 5 |
Workflow Clarity | A clearly sequenced 'Execution — 8 Steps (always run in order)' workflow with explicit input validation and a stop-on-out-of-scope gate, plus the Mandatory Output Structure acting as a checklist; the gap is that there is no explicit final self-validation/retry feedback loop before emitting the deliverable. | 4 / 5 |
Progressive Disclosure | A dedicated 'Reference Module Integration' section explicitly maps each of the eight real one-level-deep reference files to its target output section, the body serves as an orchestration/overview layer while operational rules live in references/, and all referenced paths exist in the bundle, giving clear, well-signaled navigation. | 5 / 5 |
Total | 16 / 20 Passed |