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, information-dense skill body: executable install and script commands, honest verification boundaries, and clean one-level-deep splitting of catalogs, workflows, and service contracts across reference files. It stops short of top marks on in-body example queries, inline workflow sequences, and a fully complete Resources listing.
Suggestions
Add one or two in-body example queries for the most common cases (e.g., `gget info ENSG00000141510` and a Python `gget.seq(...)` call) so the SKILL.md itself demonstrates a round-trip without requiring a hop to module_catalog.md.
Make the "### references/" Resources section list all five reference files, including module_catalog.md and common_workflows.md, so navigation does not depend on links embedded in body prose.
Tighten the longest best-practices bullets (Enrichr, Census) by moving parameter-level caveats into references/database_info.md and keeping only the actionable rule in the body.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with tool-specific facts Claude would not know (e.g., "Census `meta_only=True` returns cells, not datasets, and ignores the gene filter"; "Open Targets filters are local and applied after `limit`") with no padding of generic concepts. A few spots could be trimmed: the multi-line Enrichr best-practices bullet and the lengthy arXiv citation-fetching instructions. This fits "efficient; minor instances of over-explanation that could be trimmed" rather than the 5 anchor where every token earns its place. | 4 / 5 |
Actionability | Concrete, copy-paste-ready guidance is given: the install block (`uv venv --python 3.13 .venv` / `uv pip install "gget==0.30.8"`), exact script invocations ("`scripts/gene_analysis.py TP53`", "`scripts/enrichment_pipeline.py genes.txt --background tested_genes.txt`"), and specific calls like `gget.pdb(..., resource="mmcif")` and `gget setup <module>`. The gap versus the 5 anchor: the body itself contains no example query invocation for the most common case (e.g., a `gget info`/`gget seq` call); those examples are deferred to references/module_catalog.md. | 4 / 5 |
Workflow Clarity | Multi-step workflows are delegated to references/common_workflows.md, whose table pairs each task with an explicit checkpoint column ("Preserve Ensembl index; inspect service failures; distinguish genomic sequence from CDS"; "checksums -> downstream index"), and the Error Handling section gives a recovery pattern ("check upstream release notes and pin the newer fixed version explicitly") plus batch limits ("Process max ~1000 Ensembl IDs at once"). Not the 5 anchor because the feedback loops and full sequences live one file away rather than in the body, and a couple of workflows are summarized only as table rows. Validation for batch/paginated operations is present, so the destructive/batch cap is not triggered. | 4 / 5 |
Progressive Disclosure | The body is a genuine overview: per-parameter detail split into module_catalog.md and module_reference.md, workflows into common_workflows.md and workflows.md, service contracts into database_info.md — all one level deep, all markdown-linked, and all five referenced files verified to exist on disk. The structure sits between the 4 and 5 anchors because the "### references/" Resources section lists only three of the five files (omitting module_catalog.md and common_workflows.md, which are linked earlier in the body), a minor navigation gap against the "easy navigation" 5 anchor. | 4 / 5 |
Total | 16 / 20 Passed |