Content
65%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is genuinely actionable with accurate commands and a complete parameter table, but it is padded with generic boilerplate, has abstract workflow steps, and includes a broken reference path that weakens navigation.
Suggestions
Remove the generic scaffold sections (Lifecycle Status, Evaluation Criteria, Risk Assessment, Security Checklist, Response Template, Output Requirements) that do not add USMLE-specific guidance, or move them to a reference file.
Fix the broken `references/conditions/` reference (create the directory or remove the line) and explicitly surface guidelines.md, sample_input.json, and sample_output.json in the References section.
Tighten the Workflow to concrete case-generation steps with a validate-fix-retry loop (e.g., generate -> validate clinical plausibility against references -> regenerate) instead of abstract process language.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The core technical sections (Features, Usage, Parameters, Case Structure, Example Output) are efficient, but the body is padded with generic scaffold boilerplate (Lifecycle Status, Evaluation Criteria, Risk Assessment, Security Checklist, Response Template, Output Requirements) and vague cross-references like 'See `## Features` above for related details.' that add tokens without USMLE-specific value. | 2 / 3 |
Actionability | It provides concrete, executable commands (`python scripts/main.py --step 1 --topic cardiology --difficulty medium`, `python -m py_compile scripts/main.py`) that match the actual script args, a complete Parameters table, and a copy-paste-ready Example Output. | 3 / 3 |
Workflow Clarity | A 5-step Workflow and a Quick Check validation (py_compile) are sequenced, but the steps are generic and abstract ('validate the request', 'choose the packaged workflow') rather than concrete to case generation, with no task-specific error-recovery feedback loop. | 2 / 3 |
Progressive Disclosure | A dedicated References section signals one-level-deep files (topics.json, case_templates.json, usmle_patterns.md), and scripts/main.py exists; however the referenced `references/conditions/` directory does not exist, and several present reference files (guidelines.md, sample_input.json, sample_output.json) are not surfaced, indicating imperfect organization. | 2 / 3 |
Total | 9 / 12 Passed |