Two tools with minimal schema documentation and generic descriptions. 'digest' has a basic input schema with 3 parameters (items, top, format) but lacks output schema documentation. 'routine-prompt' has only a 'top' parameter. Both descriptions are functional but under 200 chars and lack LLM-optimized guidance on WHEN to use each tool or what downstream actions they enable. No error handling guidance, no parameter constraints beyond format enum, no idempotent/destructive hints. Schemas are present but incomplete, no documented return types, no pagination, no field-level descriptions.

• Add explicit output schema for digest: document that it returns either a markdown string (format='md') or a JSON array of scored items (format='json'), with field descriptions for each.
• Expand digest description to ~150 chars: 'Rank reading backlog items by freshness, depth, and curation signals. Returns top N items as markdown digest or JSON. Use before feeding to Claude for final triage decisions.'
• Add type and format constraints to 'items' parameter: specify it must be a JSON array of objects with required fields {url, title, saved_at} and optional {tags, snippet, word_count}. Include a link to the example schema.
• Document the scoring heuristic in digest description or as a separate 'scoring_algorithm' parameter so LLMs understand the ranking logic and can explain it to users.
• Add tool annotations: set readOnlyHint=true and idempotentHint=true for both tools to signal they are safe to retry and do not modify state.