Content
52%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 core three-step workflow is well-sequenced and actionable, but it is buried in a large amount of generic boilerplate that adds no skill-specific value and creates duplicated, self-referential sections. Trimming the template sections and adding the actual LLM prompt templates would substantially improve it.
Suggestions
Cut the generic boilerplate sections (When to Use, Key Features, Dependencies, Implementation Details, Required Inputs, Recommended Workflow, Output Contract, Validation and Safety Rules, Failure Handling, Deterministic Output Rules, Completion Checklist) — they restate process knowledge Claude already has and duplicate each other.
Merge the duplicate "Validation Shortcut" / "Quick Validation" sections and remove the "Example Usage" section that only says "See `## Usage` above"; keep one Usage section with one validation command.
Include the actual LLM prompt templates for Step 1 and Step 2 (or place them in a reference file) instead of the high-level instruction "Prompt the LLM to act as a Meta-Analysis expert", so the criteria generation is fully reproducible.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is padded with generic template boilerplate ("Validate the request against the skill boundary", "Keep the output safe, reproducible", "Required Inputs", "Completion Checklist") that restates process knowledge Claude already has. Two duplicate sections ("Validation Shortcut" and "Quick Validation") both just say run `--help`, and "Example Usage" contains only "See `## Usage` above" — several unnecessary padded sections, matching anchor 2. | 2 / 5 |
Actionability | The core workflow is concrete (the extraction command `python scripts/extract_criteria.py --inclusion ... --exclusion ...` is executable, and the `{}`-enclosure and PICO-S format requirements are specific), but the actual LLM prompt templates are absent ("Prompt the LLM to act as a Meta-Analysis expert" is direction, not instruction) and most sections by line count are abstract filler — incomplete concrete guidance, anchor 3. | 3 / 5 |
Workflow Clarity | Steps 1-3 are clearly sequenced with per-step inputs and requirements, including a consistency checkpoint (exclusions "Must NOT contradict Inclusion Criteria") and a pre-run validation shortcut. Minor validation gaps (no check that the `{}` markers were actually produced) keep it below anchor 5. | 4 / 5 |
Progressive Disclosure | Sections are labeled, script references (`scripts/extract_criteria.py`, `scripts/validate_skill.py`) are real one-level-deep paths, and no content clearly belongs in a separate file. Minor organization gaps — duplicated Usage/validation sections and the non-portable hardcoded path `cd "20260316/scientific-skills/..."` — keep it at anchor 4. | 4 / 5 |
Total | 13 / 20 Passed |