Content
77%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 well-engineered planning skill: explicit step sequence with a mandatory pre-output consistency check, a fully specified output contract, and exemplary one-level-deep reference wiring. Its main cost is token redundancy from rules restated multiple times across sections, plus the lack of a worked example to anchor output quality.
Suggestions
Deduplicate the repeated rule sets: state the self-critical risk review checklist once (Section L), the fabrication ban once (Hard Rules), the Dataset Disclaimer placement once (Section H/Step 6), and the DESeq2/limma default once (Hard Rules), then cross-reference by name from other sections.
Add a short worked example (one abbreviated filled-out section, e.g. a sample 'B. Best-Fit Study Pattern' and one row of the Section C table) so the expected output shape and depth are anchored, improving actionability.
Collapse the three overlapping 'should not' lists (Core Function, Hard Rules, What This Skill Should Not Do) into a single canonical list to reduce token cost without losing coverage.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is operational throughout — no space is spent explaining concepts Claude already knows — but it is noticeably padded by repetition: the self-critical risk review checklist appears three times (Step 7, Section L, Hard Rule 13), the Dataset Disclaimer requirement appears in at least five places, the fabrication ban in four, and the DESeq2/limma rule twice. It sits between 'mostly efficient, could be tightened' and 'several unnecessary padded sections'. | 3 / 5 |
Actionability | For an instruction-only skill the guidance is highly concrete: a fixed 7-step execution order, an exact out-of-scope redirect quote template, fully specified output sections A–L with per-section content requirements, and specific method defaults (DESeq2 for count data, limma for non-count normalized). The minor gap is the absence of a worked example or sample filled-out section showing what good output looks like. | 4 / 5 |
Workflow Clarity | The 7 steps are strictly ordered with an explicit validation checkpoint ('Step 5 — Dependency Consistency Check (mandatory before output)'), a decision gate for out-of-scope input with a ready-made redirect, a conditional Step 4.5, and a terminal self-critical risk review with fallback planning. The A–L mandatory output structure functions as a completion checklist. | 5 / 5 |
Progressive Disclosure | The SKILL.md is a clear orchestration overview, and each of the 9 reference files (all verified to exist) is explicitly mapped to the output section where it must be used ('references/study-patterns.md → Section B', etc.). References are one level deep, well signaled, and the body warns that output generated without its reference module is incomplete — matching the top anchor. | 5 / 5 |
Total | 17 / 20 Passed |