Content
48%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 workflow and prompt templates are genuinely useful and clearly sequenced, but they are buried in a large amount of generic boilerplate meta-sections that add no task-specific value. Several delivery defects (a truncated prompt, mismatched IO contract field, nonexistent CONFIG reference, contradictory validation guidance, broken heading structure) reduce actionability and navigation. Trimming the filler and fixing the prompt/IO inconsistencies would substantially raise the score.
Suggestions
Cut the generic meta-sections (When to Use, Key Features, Dependencies, Implementation Details, Recommended Workflow, Output Contract, Validation and Safety Rules, Failure Handling, Deterministic Output Rules, Completion Checklist) that restate obvious defaults, keeping only the IO Contract, Workflow, Quality Rules, and Prompts that carry task-specific content.
Fix delivery defects: complete the truncated Inclusion/Exclusion prompt, rename the `title` input to `risk_of_bias_tool` (or fix its description), remove the nonexistent `CONFIG` block reference, and reconcile the contradictory validation guidance ("Validation Shortcut" vs. "Quick Validation").
Fix the heading structure in "Prompts and Templates" so each prompt's Role/Content/Requirements are sub-headings (###) of the prompt rather than sibling top-level sections, and repair the dangling "See `## Workflow` above" reference.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Roughly half the body is generic filler that assumes no intelligence, e.g. "Execution model: validate the request, choose the packaged workflow, and produce a bounded deliverable", "Keep the output safe, reproducible, and within the documented scope at all times", and a "Completion Checklist" of self-evident confirmations. This matches anchor 2 (several unnecessary explanations or padded sections); it is not 3 because the padding is the bulk of the file rather than incidental, and not 1 because the prompts section does carry real task-specific content. | 2 / 5 |
Actionability | The six prompt templates (Search Strategy, Quality Assessment, Statistical Analysis, etc.) give concrete generation instructions, but delivery is incomplete: the Inclusion/Exclusion prompt trails off ("The user will input a section of inclusion and exclusion criteria. Please:" with nothing actionable after it), the IO contract describes input `title` as "The risk of bias assessment tool to use (e.g., ROB2, NOS, ROBINS-I...)" (a name/description mismatch), and the run plan references an in-file `CONFIG` block that does not exist. This fits anchor 3 (some concrete guidance but incomplete) rather than 4. | 3 / 5 |
Workflow Clarity | The Workflow section gives a clear 6-step sequence with per-step inputs ("Step 1: Search Strategy ... Inputs: `PICOS`, `current_time`, `language`") plus step-1 input validation and Quality Rules, matching anchor 4. It is not 5 because checkpoints are thin (no validation of `language` values, no post-compilation verification) and the body contradicts itself on validation ("Run this minimal command first ... python scripts/validate_skill.py --help" vs. "No local script validation step is required for this skill"). | 4 / 5 |
Progressive Disclosure | Everything is inlined in one 255-line SKILL.md with no references directory; inlining the six prompt templates is defensible, but organization is weak — the "Prompts and Templates" subsections use top-level `##` headings that fragment the document, "See `## Workflow` above" is a dangling cross-reference, and three overlapping validation sections (Validation Shortcut, Quick Validation, Validation and Safety Rules) obscure navigation. This matches anchor 3 (some structure but could be better organized). | 3 / 5 |
Total | 12 / 20 Passed |