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 workflow is well designed — a clear four-step sequence correctly delegating detail to a real, well-sectioned reference file and an existing helper script — but roughly a third of the body is meta-boilerplate ('Key Features', 'Implementation Details', generic 'When to Use' bullets) that wastes tokens without instructing. The end-to-end invocation of the PDF extraction script is also never shown with real arguments.
Suggestions
Delete or drastically trim the boilerplate sections ('Key Features', 'Implementation Details', 'Dependencies') and the generic 'When to Use' bullets (e.g., 'needs a packaged method instead of ad-hoc freeform output'), keeping only the task-specific trigger line.
Show a concrete end-to-end command for the helper script in the 'Helper Scripts' section, e.g. `python scripts/extract_pdf.py paper.pdf --output extracted.txt`, and drop or relativize the non-portable '20260316/scientific-skills/...' path in Example Usage.
Add an explicit validation checkpoint in Step 4, such as verifying the generated JSON parses and matches the study_risk_of_bias_schema before returning it, to close the workflow's only validation gap.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Several sections are padded boilerplate that adds nothing Claude does not already know: 'Use this skill when a data analytics task needs a packaged method instead of ad-hoc freeform output', 'Structured execution path designed to keep outputs consistent and reviewable', 'See `## Workflow` above for related details', and 'Third-party packages: not explicitly version-pinned in this skill package. Add pinned versions if this skill needs stricter environment control.' The filler spans the 'When to Use', 'Key Features', 'Dependencies', and 'Implementation Details' sections, making this noticeably verbose (anchor 2) rather than mostly efficient (anchor 3). | 2 / 5 |
Actionability | The Workflow steps are concrete (each names its prompt section in references/probast_prompts.md, all verified to exist) and the Example Usage block has runnable commands, but the body never shows a complete invocation of extract_pdf.py with an actual paper, the example cd's into a non-portable dated directory ('20260316/scientific-skills/...'), and the Helper Scripts section only says to 'use extract_pdf.py' without the actual command line. Concrete guidance exists but key execution details are incomplete, matching anchor 3; it is not 4 because the main script's usage is never demonstrated end-to-end in the body. | 3 / 5 |
Workflow Clarity | The four-step assessment sequence is clearly ordered with inputs and outputs stated per step ('Input: The risk ratings (Low/High/Unclear) from Domains 1-4', 'Output Format: A JSON object matching the study_risk_of_bias_schema'), and the run plan includes a review step. Validation is only implicit (schema conformance) rather than an explicit checkpoint, so anchor 4 fits better than 5; this is not a destructive or batch operation, so no cap applies. | 4 / 5 |
Progressive Disclosure | The body acts as an overview with one-level-deep references: every workflow step points to a named section in references/probast_prompts.md (all seven cited sections verified present in that file) and the helper script scripts/extract_pdf.py exists. Structure is good, but the 'Implementation Details' section contains a circular pointer ('See `## Workflow` above for related details') and generic bullets that muddy navigation, keeping this at anchor 4 rather than the cleanly-split anchor 5. | 4 / 5 |
Total | 13 / 20 Passed |