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.
A well-structured, navigable instruction skill with concrete search policy, a defined deliverable table, and graceful degradation rules. Its main costs are systematic repetition of the source/venue ladder across four sections and the absence of any progressive disclosure — the venue catalog and fact-extraction fields are inlined in a 290-line file rather than split into reference files.
Suggestions
State the source list and database ladder once (e.g., in Source Selection) and have Retrieval Order, External Search Policy, Step 1, and Key Rules reference it, cutting roughly 60-80 lines of repetition.
Move the Tier A/B/C venue catalog into a references/ file (e.g. references/venues.md) linked from a short 'Venue Priority' section, keeping SKILL.md as an overview.
Add a verification checkpoint after Step 2, e.g. 'If fewer than N relevant papers are found in a tier, widen to the next tier and re-search' to make the sparse-result loop explicit.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient bullet-style policy, but the source ladder is repeated four times — 'zotero, obsidian, local, ieee, sciencedirect, acm, web' appears in Source Selection defaults, Valid source values, Retrieval Order, and the Output source list, and the database ladder recurs in Retrieval Order, External Search Policy, Step 1, and Key Rules. This matches 'mostly efficient but includes some unnecessary explanation or could be tightened'; it is not a 4 because the redundancy is systematic rather than a few minor trims. | 3 / 5 |
Actionability | Guidance is concrete and executable for an instruction-only skill: exact glob paths ('papers/**/*.pdf and literature/**/*.pdf'), a defined output table with ten named columns, a fixed five-part summary order, and explicit source-parsing syntax ('$ARGUMENTS for a — sources: directive'). It stays at 4 rather than 5 because there are no copy-ready command/query templates (e.g., example Zotero MCP searches or WebSearch phrasings) for the common cases. | 4 / 5 |
Workflow Clarity | Steps are clearly sequenced (Step 0a/0b/0c → 1 → 2 → Synthesis → Output) with explicit skip conditions ('Skip this step if Zotero MCP is not configured') and degradation checkpoints ('If a source is unavailable, do not fail. Skip it silently.'). Not a 5 because there are no verification/feedback checkpoints — e.g., nothing instructs re-checking that found papers actually match the topic or re-searching when a tier comes up sparse — which are stated only as implicit 'widen if sparse' rules. | 4 / 5 |
Progressive Disclosure | This is a single ~290-line SKILL.md with no references/, scripts/, or assets/ bundle at all. Sections are well-organized and flat (one level, easy to navigate), but content that plausibly belongs in a separate reference — the ~50-line Tier A/B/C venue catalog and the field-listing for paper facts — is fully inlined, matching anchor 3 ('content that should be separate is inline'). It is above 2 because structure and headers are good and nothing is buried. | 3 / 5 |
Total | 14 / 20 Passed |