Content
75%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-structured, largely executable skill body with concrete commands, real validation checkpoints, and disciplined brevity. The main improvement opportunities are specifying the sample-sheet format, adding failure-handling feedback loops around the preflight and resource checks, and splitting the longer reference-style lists into a separate file.
Suggestions
Show the expected somatic_pairs.tsv column format (or point to an example/template file) so the runner command is fully executable end-to-end.
Add brief feedback-loop guidance for checkpoint failures — e.g. what to do when the preflight install plan is incomplete or the resource readiness check reports missing bundles.
Consider moving the resource-plan bundle details (resource_plan.json, manifest, env, readiness flags) into a reference file to keep SKILL.md closer to a lean overview.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is tight, imperative, and free of explanations of concepts Claude already knows, but a few rationale sentences in Decision Points ("A swapped or missing normal changes the biological meaning of the calls", "explicitly state the false-positive risk") could be trimmed, matching the efficient-with-minor-trimmable-instances anchor. | 4 / 5 |
Actionability | Two executable command blocks with real script paths and flags (preflight with --emit-install-plan; the Mutect2 runner with --sample-sheet, --germline-resource, --panel-of-normals) plus the sarek runner command give mostly copy-paste-ready guidance, but the somatic_pairs.tsv format is never specified and the sarek command has no example arguments, keeping it below fully-executable. | 4 / 5 |
Workflow Clarity | A clear sequence (confirm inputs → preflight → route → run → outputs) with genuine checkpoints (preflight install plan, "Verify tumor-normal pair metadata before execution", advisory-vs-blocking resource readiness) earns the clear-sequence anchor, but there are no explicit feedback loops for what to do when preflight or the resource check fails, which the top anchor requires. | 4 / 5 |
Progressive Disclosure | Sections are well organized with no dangling references and no bundle files to coordinate, but at roughly 63 lines — above the under-50-line simple-skill exception — the dense Outputs and Decision Points lists and the resource-plan paragraph could be split into a reference file, matching the good-structure-with-minor-gaps anchor. | 4 / 5 |
Total | 16 / 20 Passed |