LitLens has 5 tools with significant quality gaps. Naming is generally good (verb-noun patterns like search_arxiv_with_terms, get_paper_citations), but descriptions range from adequate to sparse. The primary tool, 'LitLens', has a verbose but thorough description (250+ chars) explaining when and why to use it. However, most other tools have minimal descriptions (20-50 chars), violating the 50-200 char baseline for LLM-optimized descriptions. Input schemas are present across all tools and properly typed (string, integer, array), but parameter descriptions are generic or missing context. For example, 'search_arxiv_with_terms' and 'search_semantic_scholar' have identical parameter definitions (query string, max_docs integer) but no guidance on differences between the two sources or when to prefer one. Output schemas are not documented in the visible code, no evidence of return type specifications. Error handling is absent; no indication of retryability, user-fixable vs fatal errors, or recovery guidance. Tool composition is reasonable (search, generate terms, get citations are distinct), but chaining metadata is unclear, e.g., does search_arxiv return paper IDs that get_paper_citations accepts? The 'sources' parameter in LitLens accepts an array of strings (['arxiv', 'semantic_scholar']), but this is an enum-like constraint that should be declared as such. Overall, the server reads as a research assistant with good intent but incomplete schema documentation and no error guidance.

• Document output schemas for all 5 tools. For search_arxiv_with_terms and search_semantic_scholar, specify: { papers: [{ id: string, title: string, authors: array, year: integer, abstract: string }], total: integer }. For get_paper_citations: { citations: [{ id: string, title: string, ... }], paper_id: string }.
• Add enum constraints to the 'sources' parameter in LitLens. Change from free-form array of strings to explicit: {type: 'array', items: {type: 'string', enum: ['arxiv', 'semantic_scholar']}, description: 'Academic sources to search. Options: arxiv (fast, open access); semantic_scholar (broader, includes paywalled papers).'}
• Expand parameter descriptions to 50-100 chars. E.g., for max_docs: 'Maximum number of papers to retrieve and analyze. Higher values are slower but more comprehensive. Default 5 for quick searches, 20 for thorough reviews.'
• Add error handling guidance to tool descriptions. E.g., search_arxiv: 'Returns an empty list if no papers match the query, try generate_related_search_terms to broaden the search.'
• Clarify paper ID compatibility. Document that search_arxiv returns arXiv IDs (e.g. '2301.12345'), search_semantic_Scholar returns Semantic Scholar IDs, and get_paper_citations expects a Semantic Scholar paper_id. If cross-source lookup is needed, add a resolve_paper_id tool.
• Add synthesis_type parameter validation to LitLens. The description lists options ('comprehensive', 'concise', 'comparative') but they should be an enum in the schema, not free-form strings. Same for sources: make it an explicit enum.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry