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 content is well-organized, concise, and gives concrete workflow guidance with validation, but its actionability and structure are undermined by the absence of the referenced kernel.py from the bundle, leaving the central executable component missing.
Suggestions
Ship the referenced kernel.py (with verify_dois, crossref_lookup, search_openalex, expand_citations, extract_dois, style_pass) in a scripts/ directory so the exec-based setup and helper calls actually work.
Re-express the retrieve → expand → synthesize → style_pass → save flow as a short numbered checklist with explicit validation checkpoints (verify DOIs, check retractions, run style_pass once) to lift workflow clarity to the top anchor.
Tighten the discursive analogy passages (e.g. 'a paragraph that hasn't been written yet') to recover tokens without losing the actionable diagnostic.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Lean and efficient text that assumes Claude's competence — it never explains what a DOI, PDF, or citation graph is — with a few discursive passages (e.g. the annotated-bibliography analogy) that could be trimmed, sitting just below the score-5 'every token earns its place' anchor. | 4 / 5 |
Actionability | Provides concrete helper names, an exact exec command, and a precise inline-citation format, but the repeatedly referenced 'kernel.py' is not present in the bundle (no scripts/ directory exists), so the core executable code the instructions depend on is missing — a key gap that leaves the guidance incomplete. | 3 / 5 |
Workflow Clarity | A clear retrieve-then-write sequence runs through the body with validation baked in (verify_dois, retraction checks, a single style_pass lint before save), but it is presented as flowing prose rather than numbered steps with explicit checkpoints, placing it just below the score-5 anchor. | 4 / 5 |
Progressive Disclosure | The body is well-sectioned and short (~55 lines) with no nested references, but it repeatedly signals a one-level reference to 'kernel.py' / 'this skill's directory' that does not exist in the bundle, so the primary referenced file is missing — a structural gap that prevents a higher score. | 3 / 5 |
Total | 14 / 20 Passed |