Server has 4 tools with basic descriptions and some parameter documentation, but significant gaps in schema completeness, parameter descriptions, and error handling. Tool names are action-oriented (search, detail, fulltext, login) which is good. However, parameter descriptions are minimal or missing for critical fields. Output schemas are not formally documented. Error handling is absent, no guidance on retries, user-fixable errors, or recovery paths. The search tool has a detailed description (>200 chars) but others are terse. No tool annotations (readOnlyHint/destructiveHint) despite clear risk profiles.

• Add formal output schema documentation for each tool. E.g. search() returns {kind: 'record'|'fulltext', title: str, author: str|null, meta: str, snippet: str, pdf_url: str|null, url: str|null, total: int|null}.
• Expand parameter descriptions with format/constraint hints. E.g. 'code: thesis identifier from search results (alphanumeric, e.g. "phd_12345")' and 'host: repository domain (e.g. "is.vsfs.cz", must be HTTPS-accessible)'.
• Add tool annotations via @mcp.tool(readOnlyHint=True) for search/detail/fulltext and destructiveHint=True for login to signal idempotency and side effects.
• Implement error classification in responses. Return {success: bool, data: {...}, error: {code: str, message: str, recovery: str}} so LLMs know whether to retry, ask user, or abort.
• Document pagination behavior: 'Returns up to 50 results per call. Use limit parameter (1-50). Response includes total count; call again with offset to fetch next page.'

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

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