Content
82%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 high-quality, actionable reference for OpenMed NER: executable examples throughout, clear parameter guidance, and a sensible privacy-first pipeline hand-off. The skill is held back only by mild redundancy in the examples and the absence of any progressive split or explicit validation checkpoints.
Suggestions
De-duplicate the output-formats example so it reuses the `note` from Quick start instead of re-importing openmed and redefining the variable.
Add one explicit verification checkpoint for the batch/JSONL path (e.g., confirm each JSONL line is valid JSON and offsets stay within `result.text` before downstream use) to lift workflow_clarity.
If the Edge cases or Key parameters sections grow further, move the parameter reference and gotchas into a references/ file and link to it from SKILL.md to deepen progressive disclosure.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Lean and assumes Claude's competence — no basic NER/library explanations — but the output-format example re-imports openmed and redefines `note`, and the Key parameters block repeats the full call signature before re-explaining three params, minor redundancy that could be trimmed. | 4 / 5 |
Actionability | Fully executable, copy-paste-ready code for quick start, every output format, JSONL saving, the CLI, and the deidentify→NER hand-off, with the common cases (dict/json/html/csv, batch JSONL, file input) all concretely covered. | 5 / 5 |
Workflow Clarity | The single analyze_text action is unambiguous and the Hand-off section sequences deidentify→NER→grounding→batch clearly, but there are no explicit validation checkpoints; this is acceptable for read-only NER yet keeps it just short of a 5. | 4 / 5 |
Progressive Disclosure | Well-organized into clearly headed sections (When to use, Quick start, Output formats, Key parameters, CLI, Hand-off, Edge cases, Standards & references) with one-level-deep external links, but with no bundle files everything lives inline in ~190 lines, above the 50-line simple-skill threshold that would justify a 5. | 4 / 5 |
Total | 17 / 20 Passed |