Content
31%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 dominated by generic genomics textbook content and duplicated template boilerplate rather than lean, skill-specific instruction, and its code examples do not fully match the bundled placeholder script. Progressive disclosure is inverted: heavy inline reference material while the one existing reference file is mislabeled as absent.
Suggestions
Move Common Patterns, Common Pitfalls, Troubleshooting, and the quality checklists into files under references/ and link them from a short overview; also fix the false claim that no reference files exist (references/runtime_checklist.md is present).
Cut the duplicated boilerplate sections (When to Use / Key Features / Dependencies / Example Usage / Implementation Details overlap) and repair the broken "See ## Prerequisites/Usage/Workflow above" cross-references and out-of-order sections.
Align examples with the actual script: define or remove the undefined identify_somatic call, disclose prominently that call_cnvs/plot_genome_wide are placeholders (or implement them), and replace commented-out '# Example invocation' lines with real, runnable commands using actual file inputs.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~1180-line body extensively explains domain concepts Claude already knows — "Somatic vs Germline Classification" tables, a full "BED Format Specification", "CNV Detection Methods Comparison", pages of "Common Pitfalls" and "Troubleshooting" for standard genomics issues — and is heavily padded with duplicated boilerplate sections ("See `## Prerequisites` above", "See `## Usage` above", "See `## Workflow` above") for a script that is 68 lines. This matches the 'severely verbose; extensively explains concepts Claude already knows; heavily padded' anchor. | 1 / 5 |
Actionability | There is real, executable-style guidance: concrete Python API examples ("caller.call_cnvs(...)", "caller.save_bed(...)"), CLI commands whose flags match the actual script (--input, --reference, --output, --bin-size, --plot-format), and a runnable quick check. But guidance is incomplete: the "Complete Workflow Example" calls an undefined "identify_somatic" function, "plot_genome_wide" in the bundled script only prints a path without creating a plot (undisclosed except as a buried 'Known Issues' note), and example invocations are commented-out placeholders with nonsense inputs ("--input 'Audit validation sample with explicit symptoms...'"), matching the 'some concrete guidance but incomplete' anchor. | 3 / 5 |
Workflow Clarity | Sequences and checkpoints exist — a 5-step "## Workflow" with fallback, a 4-step "Complete Workflow Example", and pre/during/post-analysis checklists with critical items. But the workflow is fragmented across at least four overlapping sections (Workflow, Example Usage run plan, Implementation Details, Complete Workflow Example) with broken "See ## X above" cross-references and out-of-order sections (Usage and Prerequisites appear near the end), so checkpoints are scattered rather than clearly sequenced — between the 'validation gaps' (3) and 'most checkpoints present' (4) anchors, landing at 3. | 3 / 5 |
Progressive Disclosure | The body claims "(No reference files currently available for this skill)" while references/runtime_checklist.md actually exists and is never linked, and "Reference material available in `references/`" is contradicted one section later. Meanwhile ~900 lines of reference-grade material (Common Patterns, Quality Checklist, Common Pitfalls, Troubleshooting, method comparisons) that clearly belongs in separate reference files are inlined in SKILL.md, matching the 'content that clearly belongs in separate files is inlined' anchor; structure via headers keeps it above 1. | 2 / 5 |
Total | 9 / 20 Passed |