Conciseness

The body is lean and example-driven — each subcommand is a one-line description plus a bash snippet, with no explanations of biology or libraries Claude already knows. Minor trimmable redundancy exists: the Overview's 10-item bullet list repeats the per-section descriptions, and NCBI_API_KEY guidance appears in both Prerequisites ("Raises the NCBI rate limit from 3 to 10 requests/second") and Core Rules ("the query speed limits are automatically increased significantly"). Not 5: those duplications could be cut; not 3: there is no padded or unnecessary explanation beyond these minor instances.

Actionability

Nearly every subcommand has a copy-paste-ready command with realistic arguments (e.g. `uv run scripts/ncbi_fetch.py gene-protein WRR4B --organism "Arabidopsis thaliana"`), and output handling is specified ("All subcommands write structured JSON output. Use `--output FILE`..."). The gap: the Patent Sequence Search examples lost their code fences ("`bash uv run scripts/ncbi_fetch.py patent-search --patent-number US10123456...`"), rendering section 9's commands as broken inline text. Not 5: one section's examples are malformed; not 3: guidance is genuinely executable throughout.

Workflow Clarity

The "Standard Sequence Retrieval Cascade" gives an explicit 7-step priority order, and result interpretation is addressed ("When multiple results are returned, select by: Closest match to expected length... Source priority (RefSeq > GenPept > patent)"), plus an error-recovery hint ("the tool will report `is_genomic: true` so you can fall back to a homology-based approach"). Not 5: there is no explicit validate-then-proceed checkpoint (e.g. verifying fetched length against expectation is only implied by --target-length); not 3: the sequence is explicit with most checkpoints present, and operations are read-only so no destructive-batch cap applies.

Progressive Disclosure

scripts/ncbi_fetch.py is properly referenced and exists, and section structure is clear, but references/citation.bib is present in the bundle yet never mentioned or linked anywhere in the body — an orphaned, unnavigable reference. Additionally, the full 10-subcommand command reference (~150 lines) is inlined in SKILL.md, which is heavier than an overview pointing to detailed materials. Not 4: an unlisted bundle file is more than a minor organization gap; not 2: the structure and headings are good and paths that are referenced resolve correctly.

Specificity

"Retrieve protein and nucleotide sequences from NCBI databases using E-utilities" names the domain and primary action, then enumerates seven concrete operations ("direct accession lookup, CDS translation, gene+organism search, locus lookup, PubMed-linked sequences, patent protein extraction, and organism+length fallback search") — comprehensive coverage of the skill's retrieval workflows, matching the anchor for multiple specific concrete actions. Not 4: the enumerated actions cover the skill's workflows comprehensively rather than leaving minor gaps.

Completeness

The "what" is explicit ("Retrieve protein and nucleotide sequences from NCBI databases... Supports direct accession lookup, CDS translation...") and the "when" is explicit with concrete triggers ("Use when you need to fetch biological sequences by accession, gene name, locus tag, PubMed ID, or patent number"). This clearly matches the anchor for answering both what AND when with concrete trigger phrases.

Trigger Term Quality

Natural user phrases are well covered: "accession", "gene name", "locus tag", "PubMed ID", "patent number", "protein", "nucleotide", "organism". However common domain terms users actually say — "FASTA", "Entrez", "GenBank", "RefSeq", "DNA sequence" — are absent, so it is good-but-not-comprehensive keyword coverage. Not 3: most natural trigger phrasing for this domain is present; not 5: no synonyms like FASTA/.fasta are included.

Distinctiveness Conflict Risk

"from NCBI databases" scopes the what clearly to a niche, but the trigger clause "fetch biological sequences by accession" does not re-mention NCBI, so a bare accession (formats shared with UniProt/Ensembl, which have sibling skills per the body) carries minor overlap risk. Not 5: the when-clause triggers are not fully disambiguated from sibling database skills; not 3: the NCBI/E-utilities framing makes it mostly distinct.