Content
85%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A concise, actionable overview with excellent progressive disclosure and real reference files. The main gap is workflow clarity: it gives principles and tool picks rather than an end-to-end sequenced pipeline with validation checkpoints.
Suggestions
Add a short ordered canonical pipeline (e.g., FASTQ → FastQC/MultiQC → align → mark-duplicates → call → annotate) with explicit PASS/FAIL validation gates between stages.
Include one concrete QC threshold example (read depth, mapping rate, contamination) so the 'set clear PASS/FAIL thresholds' principle becomes actionable.
Show a one-line resume/checkpointing command or flag, since reproducibility and resume are named as principles but not demonstrated.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Lean, token-efficient body — terse opinionated defaults ('Use BWA-MEM2', 'prefer CRAM over BAM (30-50% smaller)') that assume Claude's competence and avoid explaining what FASTQ/BAM/VCF are beyond a one-line role label. | 3 / 3 |
Actionability | Concrete executable guidance throughout: specific tool names per task, a runnable nf-core command with flags, and concrete file-format conventions (bgzip + tabix, .bai/.crai/.tbi/.csi indexing). | 3 / 3 |
Workflow Clarity | Pipeline-design principles are listed and workflow types are routed to references, but there is no explicit sequenced pipeline procedure with validation checkpoints (e.g., FASTQ→QC→align→call→annotate with PASS/FAIL gates); the PASS/FAIL thresholds are mentioned as a principle but not wired into a step sequence. | 2 / 3 |
Progressive Disclosure | Clear overview that splits detail into one-level-deep, well-signaled references (references/rnaseq.md, references/annotation.md, references/cnv.md), all of which exist; the nf-core table keeps the overview scannable. | 3 / 3 |
Total | 11 / 12 Passed |