Content
87%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is a tight, actionable overview that defers detail to well-signaled reference files and relies on Claude's domain knowledge. Its main gap is the absence of an explicit validation feedback loop in the workflow guidance for batch pipeline operations.
Suggestions
Add a short sequenced workflow (e.g. QC reads → align → call → annotate) with an explicit validate/verify checkpoint and a fix-and-retry loop, since pipeline runs are batch operations.
State concrete PASS/FAIL QC thresholds or point to where they are defined, rather than only mentioning that thresholds should be set.
Clarify how to choose between nf-core reuse and custom pipeline construction (a one-line decision rule) so the guidance is more prescriptive.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Lean bullet lists of opinionated tool choices and formats assume Claude's existing knowledge (e.g. "FASTQ: Raw reads", "prefer CRAM over BAM (30-50% smaller)"); no padding or explanation of basic concepts. | 3 / 3 |
Actionability | Concrete, specific guidance throughout (BWA-MEM2, STAR, DeepVariant, Mutect2, VEP) plus a copy-paste-ready nf-core example command; actionable even though most content is recommendation rather than code. | 3 / 3 |
Workflow Clarity | Workflow types are enumerated and QC/PASS-FAIL thresholds and file validation are mentioned, but there is no explicit sequenced pipeline with validate→fix→retry checkpoints; for batch-oriented bioinformatics operations this caps the score per the rubric. | 2 / 3 |
Progressive Disclosure | SKILL.md is a concise overview that cleanly signals one-level-deep references (rnaseq.md, annotation.md, cnv.md), all of which exist as real bundle files; content is appropriately split. | 3 / 3 |
Total | 11 / 12 Passed |