Content
93%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.
Excellent content: fully executable guidance, dense with non-obvious operational knowledge (download dir, ESM VRAM behavior, output refusal on non-empty dirs) and a well-organized error table. The only gap is the absence of an explicit validation loop for batched multi-call campaigns.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes Claude's competence: positioning versus Boltz-2/AlphaFold3, licensing, an executable snippet, format spec, and two high-value gotchas with no filler or explanation of concepts Claude already knows. | 5 / 5 |
Actionability | It provides a complete, copy-paste-ready run_inference call, exact FASTA header syntax ('>{entity_type}|name={id}'), the CLI equivalent ('chai-lab fold complex.fasta out/ --use-msa-server'), output filenames, score fields, and a ranking rule with an ipTM threshold — covering the common cases fully. | 5 / 5 |
Workflow Clarity | The sequence (write FASTA, run, rank by aggregate_score with iptm > 0.5 as a soft pass, recognize errors, proceed to filtering) is clear and backed by an error-recovery table, but there is no explicit validate-then-proceed feedback loop for the batched campaigns the description invites, so it falls short of the 5 anchor. | 4 / 5 |
Progressive Disclosure | No bundle files exist and none are needed; the body is compact and cleanly sectioned (running, two gotcha sections, error table, next-step pointer) with no dangling references, satisfying the simple-skill standard for full marks. | 5 / 5 |
Total | 19 / 20 Passed |