Content
67%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 operative core is strong — a verified, executable script command with matching CLI flags, complete classification/summarization prompts, a specified output format, and quality rules guarding against hallucination and empty results. However, roughly a third of the body is auto-generated template padding that repeats the description verbatim three times, includes a non-portable dated example path, and contains a broken cross-reference ('See ## Workflow above'). Cutting the When to Use/Key Features/Dependencies/Implementation Details boilerplate would make this a substantially leaner, higher-scoring skill.
Suggestions
Conciseness: delete the template filler — collapse 'When to Use', 'Key Features', 'Dependencies', and 'Implementation Details' into a short 'When to Use' and a brief dependency note, removing the three verbatim repeats of the frontmatter description.
Conciseness/actionability: replace the non-portable Example Usage path ('20260316/scientific-skills/...') with a repo-relative example and a realistic keyword (e.g., 'python scripts/pubmed_search.py --keywords "lung cancer" --days 7 --limit 20 --format markdown').
Workflow clarity: add a validation checkpoint for script failures — check the exit status and handle API/network errors or malformed JSON output, not just the empty-results case; also fix the broken 'See ## Workflow above' cross-reference.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The core Workflow, prompts, and Quality Rules sections are lean, but roughly a third of the body is template padding: the description is repeated verbatim three times ('Scope-focused workflow aligned to: Generates a weekly academic literature report...'), 'Key Features' and 'Dependencies' ('not explicitly version-pinned in this skill package. Add pinned versions if...') are filler, and 'Implementation Details' restates generic best practices Claude already knows ('validate the request, choose the packaged workflow'). Not 2 because the operative ~70% (workflow commands, prompts, output format, error handling) has no over-explanation and every token there earns its place. | 3 / 5 |
Actionability | The workflow provides a copy-paste-ready command ('python scripts/pubmed_search.py --keywords "{keywords}" --days 7 --limit 20 --format markdown') whose flags match the actual bundled script, plus complete classification and summarization prompts with a JSON output contract and a concrete empty-results message. Not 5 because the Example Usage section cd's into a non-portable dated path ('20260316/scientific-skills/Evidence Insight/research-article-weekly') and the run plan includes hedged, non-executable steps ('Edit the in-file CONFIG block or documented parameters if...'). | 4 / 5 |
Workflow Clarity | The four-step sequence (draft report via script → optional AI refinement with classification prompt → grouped final report → combined output with Executive Summary) is clearly ordered, and Quality Rules supply checkpoints: 'Must use real data returned from the pubmed_search.py script. Do not hallucinate papers', 'Ensure all retrieved and relevant papers are included', and explicit no-results handling. Not 5 (capped below) because there is no validation for script-level failures — network/API errors, malformed output, or non-zero exit — only the empty-results case is handled. | 4 / 5 |
Progressive Disclosure | Structure is good: the single bundle file 'scripts/pubmed_search.py' is referenced by exact path and exists, headers organize the body, and the inline AI-refinement prompts are appropriately placed in SKILL.md as core operating instructions rather than pushed to nested references. Not 5 because of minor organization gaps: 'See `## Workflow` above for related details' points the wrong way (Workflow appears below that section), and the template filler sections (Key Features, Dependencies, Implementation Details) dilute the overview's navigability. | 4 / 5 |
Total | 15 / 20 Passed |