Content
70%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-sequenced, actionable workflow with strong error recovery and interpretation guidance, but it is a long single-file skill with systematic caveat repetition and no progressive disclosure into reference files. Tightening duplicated caveats and splitting optional tool detail (DTU, Boltz) into references would materially improve it.
Suggestions
State each caveat once — collapse the repeated ~800-residue ESMFold limit and single-chain limitation into the Limitations section (or their primary phase) and reference it elsewhere.
Move optional tool detail — the DTU model table and Boltz paid-workflow section — into reference files (e.g., references/dtu-tools.md, references/boltz.md) linked from short overview paragraphs.
Add one concrete example invocation and a short parsing snippet (e.g., computing mean pLDDT and low-confidence ranges from the returned array) to make the Phase 2 workflow copy-paste executable.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The bulk is genuinely useful tool-specific detail (parameters, thresholds, report templates), but key caveats are systematically repeated: the ~800-residue ESMFold limit appears six times (KEY PRINCIPLES, Input Parameters, Phase 0, Fallback Strategies, Databases, Limitations) and the single-chain limitation four times, and KEY PRINCIPLES duplicates the phase content. | 3 / 5 |
Actionability | Concrete guidance throughout: exact tool names, parameter formats with examples ("P04637 R175H"), quantitative pLDDT/pTM interpretation tables, a report template, and per-failure fallbacks. As an instruction-only skill, absent code is acceptable, but there are no example invocations or parsing snippets (e.g., computing mean pLDDT from the returned array is described rather than shown), leaving minor gaps. | 4 / 5 |
Workflow Clarity | The six-phase pipeline is explicitly sequenced with an overview diagram, per-phase step lists, decision logic, a Fallback Strategies table covering each failure mode (error-recovery loops), and a Quality Minimums checklist — matching the top anchor. Operations are read-only/computational, so the destructive-batch cap does not apply. | 5 / 5 |
Progressive Disclosure | No bundle files exist and all ~370 lines are inline in SKILL.md. Internal structure is well organized with clear headers, but detail that plausibly belongs in reference files (DTU model table, Boltz paid-workflow specifics, tool parameter reference) is fully inlined with no pointer files, so the skill is a monolithic single-file overview. | 3 / 5 |
Total | 15 / 20 Passed |