Content
50%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 domain-rich, largely actionable skill document with real executable code, CLI examples, and a complete output schema, but it is noticeably verbose with generic boilerplate sections and inlines bulk reference material that would be better split into named files. Workflow validation checkpoints are present at the edges but underdeveloped in the core prediction path.
Suggestions
Trim or remove generic boilerplate sections ('Lifecycle Status', 'Risk Assessment', 'Evaluation Criteria', generic 'Workflow'/'Output Requirements') that restate process Claude already knows, to recover token budget.
Move the large HLA support list, scoring-algorithm details, and full output schema into reference files (e.g., references/hla_alleles.md, references/scoring.md) and link to them by name from the body.
Add an explicit validation feedback loop to the core prediction workflow (run py_compile, execute on sample mutations, verify output against expected schema, fix and re-run) rather than leaving checkpoints implicit.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body runs ~450 lines and includes several padded/generic boilerplate sections ('Lifecycle Status', 'Risk Assessment', 'Evaluation Criteria', generic 'Workflow', 'Output Requirements') that explain process Claude already knows, alongside dense useful domain content, matching the 'noticeably verbose; several unnecessary explanations or padded sections' anchor. | 2 / 5 |
Actionability | It provides concrete, mostly executable guidance — an importable Python API example (NeoantigenPredictor class, real in scripts/main.py), CLI invocations with real flags, a full output JSON schema, and weighted scoring formulae — with only minor gaps such as the symbolic 'Σ(wi × fi)' immunogenicity expression, fitting the 'mostly executable guidance with minor gaps' anchor. | 4 / 5 |
Workflow Clarity | A sequenced 'Workflow' with Quick Check (py_compile) and Audit-Ready Commands provides some checkpoints, but the core prediction workflow lacks explicit validation feedback loops (validate -> fix -> retry) for the multi-step analysis, sitting at the 'steps listed but validation gaps; checkpoints implicit' anchor. | 3 / 5 |
Progressive Disclosure | Sections are well-headed and references are one level deep with real files (references/README.md, example_mutations.csv, scripts/main.py verified to exist), but large reference-style blocks (HLA support list, scoring algorithms, output schema) are inlined and references are signaled generically ('See references/ directory') rather than by named file, matching the 'some structure but could be better organized; references not clearly signaled' anchor. | 3 / 5 |
Total | 12 / 20 Passed |