Scoring was not performed

• Convert this to a proper MCP server using mcp-python SDK or similar. Implement an MCP-compliant server with tool registration, JSON Schema input definitions, and protocol message handlers.
• Define formal JSON Schema for all tool inputs. For tools with no visible parameters (generate_rss, generate_sitemap, update_daily_anthropic), add schema definitions documenting expected inputs, defaults, and constraints.
• Document output schemas for every tool. Specify what fields are returned, their types, and how they chain to downstream tools (e.g., what IDs or references are included).
• Add actionable error handling. Instead of raw exceptions, return structured error messages: error code, human-readable message, recovery suggestion, and list of alternatives when applicable.
• Implement dry-run and confirmation patterns for destructive operations (check_changelog writes to CHANGELOG.md, generate_digest prepends to NEWS.md, email_digest sends emails). Make these reversible or add undo tools.
• For summarize_news_item and batch_summarize, inject ANTHROPIC_API_KEY via secure MCP context mechanism, not direct environment variable access. Ensure secrets never appear in tool parameters or logs.
• Add idempotency documentation for all tools with side effects. Specify whether calling twice with same input produces duplicate side effects or is safely idempotent.
• For batch_summarize, document the array schema of 'items' parameter: what fields (title, url, body) are required vs optional, and what types are expected.