Conciseness

The body is dense and table-driven with no explanation of concepts Claude already knows, but a few spots could be trimmed — e.g. the remote-compute prose "A final .result() read reports whether its follow-up was suppressed or had already been committed; otherwise the app starts the later analysis turn for an unread final result" is wordy and confusing. Not score 5 because these minor instances of over-writing remain; not score 3 because they are isolated, not a pattern of padding.

Actionability

Fully executable, copy-paste-ready code throughout: install ("pip install evo2"), loading/scoring (Evo2("evo2_7b") with score_sequences), generation with all parameters, a concrete submitJob call, and troubleshooting rows with exact fixes ("Set HF_HUB_OFFLINE=1", "Pass list[str]"). Common cases (7B scoring, generation, remote jobs) are all covered.

Workflow Clarity

A clear sequence is present (prerequisites → install → load/score → generate → remote compute), and the remote-compute flow has explicit checkpoints (read compute_details, submit, retain job_id, query with attachJob), plus a troubleshooting table as a feedback loop. Not score 5 because there is no explicit validate-then-proceed step for scoring runs (e.g. verifying job status before consuming outputs), leaving minor validation gaps.

Progressive Disclosure

No bundle files exist, and the single SKILL.md is well-organized into tight sections (Prerequisites, How to run, Models, Output format, Decision tree, Remote compute, Performance, Troubleshooting) with the only external pointers — the remote-compute-ssh skill and compute_details — clearly signaled and one level deep. Content is appropriately placed in one file; nothing that belongs elsewhere is inlined and nothing is nested.

Specificity

"Score, embed, and generate DNA sequences" plus four concrete numbered actions — "Computing per-nucleotide or per-sequence likelihoods for variant effect scoring", "Embedding genomic windows for downstream classification", "Generating DNA conditioned on a prefix", "Scoring regulatory or coding regions across species" — is comprehensive, multi-action coverage with no vague filler.

Completeness

"Score, embed, and generate DNA sequences with Evo 2, a long-context genomic foundation model" states the what, and "Use this skill when: (1)…(4)" explicitly enumerates the when with concrete trigger cases. Not score 4 because the when-clause is fully explicit and enumerated, not merely present.

Trigger Term Quality

Natural phrases users would say are present ("DNA sequences", "variant effect", "generate DNA", "embedding"), but common synonyms and variations are missing — e.g. "mutation", "variant scoring", "DNA language model", "log-likelihood of a sequence" — so coverage is good rather than exhaustive.

Distinctiveness Conflict Risk

The niche is narrow (a DNA genomic foundation model for likelihoods/embeddings/generation) and Evo 2 is named directly, so it is clearly distinguishable from generic data or document skills. Not score 4 because the only conceivable overlap is with other genomic models (e.g. track predictors), and the description's likelihood/embedding/generate framing separates those cleanly.