Content
92%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 lean, dense instruction-only skill that assumes Claude's competence and delivers only harness-specific facts: exact MCP tool names, a tool-naming pattern, key requirements, fallback chains, and a clear three-step workflow. The main gap is the absence of any example tool invocation or parameter guidance, which keeps actionability at 4.
Suggestions
Add one short example call (e.g. search_papers with a query and source filter) or a parameters note for the core tools so actionability reaches copy-paste-ready.
Clarify "when the field is clear" in the targeted-search tip — rephrase as 'when you know the venue or index (e.g. arXiv preprint, biomedical), use the specific search_<source> tool'.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Every sentence carries non-obvious operational facts Claude could not know — tool names, the source naming pattern, IEEE key requirement, Scopus full_text semantics, cache behavior, config paths, and bot-blocking fallbacks. There is no concept explanation or padding, fitting the "every token earns its place" anchor. Not score 4 because there is no identifiable over-explanation to trim. | 5 / 5 |
Actionability | Guidance names exact tools ("mcp__paper-search__search_papers", "download_with_fallback", the search_<source>/read_<source>_paper pattern) and gives concrete output specs ("a table: title, authors, year, source, DOI/URL", "report the saved path"). It falls short of fully executable copy-paste-ready guidance because no example invocation or parameter sketch for any tool is shown — a minor gap, matching the score-4 anchor. | 4 / 5 |
Workflow Clarity | The three-step Workflow section (search → present results table → read/download with fallback) is unambiguous, and failure handling is addressed elsewhere in the body ("fall back to other sources rather than retrying", respect Scopus full_text/abstract_only/unavailable status). This is a read-only, single-purpose skill, so the simple-skill exception applies and no destructive-operation validation cap does. | 5 / 5 |
Progressive Disclosure | The skill is under 50 lines with no bundle files (no references/, scripts/, or assets/ exist) and no need for external references; per the scoring notes, well-organized sections alone qualify for 5 here. The body is cleanly sectioned (Core tools, Source-specific tools, Workflow, Notes) with no inlined content that belongs in a separate file. | 5 / 5 |
Total | 19 / 20 Passed |