Content
63%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 body delivers a well-sequenced, mostly executable multi-source workflow with strong graceful-degradation behavior. Its weaknesses are repeated content (the sources-directive examples and de-duplication rules appear multiple times) and a monolithic single-file layout that inlines per-source procedures which would be better split into reference files.
Suggestions
Remove the duplicate `— sources:` example blocks: the Overrides callout (lines 26-33) and the Source Selection examples (lines 45-55) list the same directives — keep one and cross-reference it.
Split the per-source procedures (Zotero Step 0a, Obsidian Step 0b, the Semantic Scholar integration, and the Step 6 research-wiki workflow) into reference files under references/ and keep SKILL.md as a concise overview with clearly signaled one-level-deep links.
Consolidate the three separate de-duplication instructions (Step 0c item 2, Step 1, and the arXiv/S2 section) into a single de-duplication rule stated once, and add an explicit validate-and-retry loop for the batch PDF-download step (e.g., re-try failed downloads once before reporting).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is operational rather than tutorial, but the `— sources:` directive examples appear twice in near-verbatim form (the Overrides block and the Source Selection examples block), and de-duplication guidance is repeated in Step 0c, Step 1, and the arXiv/S2 section. Matches the 3 anchor (mostly efficient but could be tightened); not 4 because the duplicated blocks are a clear trim of ~15 lines, and not 2 because nothing explains concepts Claude already knows. | 3 / 5 |
Actionability | Quotes concrete executable guidance: `python3 "$SCRIPT" search "QUERY" --max 10`, `--fields-of-study "Computer Science,Engineering" --publication-types "JournalArticle,Conference"`, `python3 "$SCRIPT" download ARXIV_ID --dir papers/`, and the research_wiki.py commands, plus find-based script discovery with explicit fallbacks. Mostly executable with minor gaps (unresolved "QUERY" placeholders, undefined MCP tool names); not 5 because copy-paste readiness is broken by those placeholders and the 'try common patterns and adapt' MCP guidance. | 4 / 5 |
Workflow Clarity | Quotes a clear sequenced workflow (Step 0a → 0b → 0c → 1 → 6) with explicit skip conditions ("Skip this step entirely if Zotero MCP is not configured") and graceful degradation ("If `arxiv_fetch.py` is not found, fall back to WebSearch"). Downloads include a checkpoint ("Verify each PDF > 10 KB"); not 5 because there is no validate→fix→retry feedback loop, though the batch download step does carry per-item verification so the destructive/batch cap at 3 does not apply. | 4 / 5 |
Progressive Disclosure | No bundle files exist (no references/, scripts/, or assets/ directories), so all 236 lines live inline in SKILL.md; per-source detail such as the full Zotero procedure (Step 0a), Obsidian procedure (Step 0b), Semantic Scholar integration, and the research-wiki step are candidates for separate reference files. Matches the 3 anchor ("content that should be separate is inline") despite good section headers; not 4 because the length and per-source depth would clearly benefit from splitting, and not 2 because section structure and navigation are genuinely present. | 3 / 5 |
Total | 14 / 20 Passed |