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.
The body is a well-structured, executable reference: every subcommand has concrete commands, a clear retrieval cascade orders the workflows, and result-selection criteria are spelled out. Weaknesses are cosmetic but real — duplicated subcommand listings, two broken code/markdown blocks in sections 3 and 9, and an orphaned bundle file (references/citation.bib) that is never surfaced to the reader.
Suggestions
Fix the malformed code fences in the Patent Sequence Search section (the two `uv run ... patent-search` examples are inline text, not fenced bash blocks) and the broken numbered list in the CDS Translate section so all examples are copy-paste executable.
Eliminate the duplication between the Overview's 10-item subcommand list and the Utility Scripts section — keep the overview as a one-line index of names only, or drop it and let the section headers serve as the index.
Reference references/citation.bib in the body (e.g., a line in the Reference or Core Rules section noting when to cite the Sayers 2025 E-utilities paper), since it is a provided bundle file that is currently undiscoverable.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is largely tight — each subcommand gets one descriptive line plus executable commands, and it never explains concepts Claude already knows (no 'what is NCBI' padding). Minor inefficiencies: the Overview re-lists all 10 subcommands that the Utility Scripts section then re-describes, and the Prerequisites section runs long. Fits 'efficient; minor instances that could be trimmed' rather than the every-token-earns-its-place anchor. | 4 / 5 |
Actionability | Nearly all guidance is copy-paste ready: 'uv run scripts/ncbi_fetch.py fetch-protein XP_022033624 -o /tmp/result.json' and similar concrete invocations with real flags for every subcommand. Section 9's patent examples are mangled ('`bash uv run scripts/ncbi_fetch.py patent-search --patent-number US10123456...' rendered as inline code instead of fenced blocks) and section 3's numbered method list is broken markdown, so a couple of examples are not cleanly executable — minor gaps, not missing steps. | 4 / 5 |
Workflow Clarity | A clear 7-step 'Standard Sequence Retrieval Cascade' sequences the subcommands by priority, and 'Interpreting Results' gives explicit selection criteria (target_length match, header relevance, 'RefSeq > GenPept > patent' source priority) plus fallback guidance ('If the accession is a genomic record... fall back to a homology-based approach'). It stops short of explicit verify checkpoints per step (e.g., 'confirm returned length matches expected before returning'), matching 'clear sequence with most checkpoints present; minor validation gaps'. | 4 / 5 |
Progressive Disclosure | Good structure: the body clearly signals 'scripts/ncbi_fetch.py — Single script with subcommands' (a real, verified bundle file) and organizes the 10 subcommands under headed sections with an overview index. However, the bundle's references/citation.bib (the Sayers 2025 E-utilities citation) is never mentioned anywhere in the body — an unreferenced bundle file — a minor organization/navigation gap that keeps it below the well-signaled-references anchor. | 4 / 5 |
Total | 16 / 20 Passed |