Content
42%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 reads like template-generated boilerplate stitched around a real domain core: the description is duplicated twice, generic governance sections pad the token budget, and one bullet is leftover from an academic-writing template. Most critically, the showcase CLI and API examples reference a non-existent script path, non-existent CLI flags, and a non-existent create_summary() method, so a user following them verbatim would immediately fail.
Suggestions
Fix the executable examples to match the actual bundle: replace 'scripts/discharge_drafter.py' with 'scripts/main.py', correct the import to 'from scripts.main import DischargeDrafter', remove the undocumented CLI flags (--batch, --priority, --auto-escalate-risk, --mode, --reading-level, --language, --output), and remove or implement the non-existent drafter.create_summary() call.
Cut the padded boilerplate: delete the two verbatim re-pastes of the frontmatter description ('When to Use', 'Key Features') and the generic 'Implementation Details'/'Output Requirements'/'Response Template'/'Input Validation' sections, and drop the mismatched 'academic writing tasks' bullet.
Move the bulk API reference ('Core Capabilities', 'Common Patterns', quality checklists) into references/ files and link them by name (e.g., discharge_template.md) so SKILL.md stays a lean overview with clearly signaled one-level-deep references.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Noticeably verbose with several padded sections: the frontmatter description is pasted verbatim twice in the body ('When to Use' line and 'Key Features' line), and generic governance boilerplate ('Implementation Details', 'Output Requirements', 'Response Template', 'Input Validation') adds little skill-specific value. Line 16 even instructs 'Use this skill for academic writing tasks' — mismatched template residue from a different skill. Not 1 because the body does not explain basic concepts Claude already knows and contains real domain content (risk stratification tables, QA checks); not 3 because duplicated description text, four boilerplate sections, and a mismatched bullet go beyond isolated tightening. | 2 / 5 |
Actionability | Concrete code and commands exist, but a significant fraction will not execute against the actual bundle: the CLI section targets 'scripts/discharge_drafter.py' which does not exist (only scripts/main.py), and shows flags (--batch, --priority, --auto-escalate-risk, --mode, --reading-level, --language, --output) that main.py does not define (it accepts only --patient, --digital-twin-model, --include-predictions, --output-format, --output-dir). The Quick Start imports 'from scripts.discharge_drafter import DischargeDrafter' (wrong module) and Core Capability 1 calls 'drafter.create_summary()', which does not exist in main.py. Not 4 because these gaps are not minor — the primary CLI and API examples fail as written; not 2 because some examples are genuinely correct (generate(), simulate_outcomes() signatures match, and the py_compile/--help commands target the real script). | 3 / 5 |
Workflow Clarity | The 'Workflow' section lists a coherent 5-step sequence (confirm -> validate scope -> execute -> structured result -> fallback) plus a 'Quick Check' py_compile validation, but the steps are generic template prose not tied to concrete commands or checkpoints per step. The batch CLI example ('--batch high_risk_patients.csv') involves batch operations with no validation step, which per the rubric caps this dimension at 3. Not 2 because a real sequence with error-handling and fallback guidance is present; not 4 because checkpoints are implicit and the batch path lacks any verification step. | 3 / 5 |
Progressive Disclosure | The bundle provides references/discharge_template.md and references/medical_terms.json, but the body only gestures at 'Reference material available in `references/`' without naming or linking either file. Meanwhile ~100 lines of API reference ('Core Capabilities', 'Common Patterns', checklists) that belong in reference files are inlined in SKILL.md. Not 2 because the document is well-sectioned with headers and references do exist; not 4 because references are not clearly signaled by name and bulk API content is inline rather than split out. | 3 / 5 |
Total | 11 / 20 Passed |