AgentClip provides 7 tools with complete JSON schemas and basic descriptions. Naming follows verb_noun convention (store_, search_, get_, list_, update_, delete_). However, descriptions are terse (avg 60 chars, well below the 194-char baseline), lack WHEN/WHY context, and omit parameter-level documentation. No output schemas documented. Error handling is absent, no recovery guidance, no actionable error messages. Parameter descriptions are minimal or missing entirely (e.g., 'limit' has no range constraints). No tool annotations (readOnlyHint, destructiveHint). Schemas are present but sparse, missing descriptions for most parameters.

• Expand tool descriptions to 100 - 200 chars. Include WHAT (action), WHEN (use case), and WHAT IT RETURNS. E.g., 'Store a text snippet to your AgentClip library. Use this to save articles, code, notes, or quotes for later retrieval. Returns the new snippet ID and metadata.'
• Add parameter descriptions for all inputs. For 'limit', specify range: 'Maximum number of results to return (1 - 100, default 20).' For 'query', add: 'Search term or phrase (min 1 char, max 500 chars).'
• Document output schemas for each tool. E.g., store_snippet returns {id: number, content: string, title?: string, created_at: ISO8601, tags?: string[]}. This enables LLMs to chain tools correctly.
• Add tool annotations: mark search_snippets, semantic_search, get_snippet, list_recent with readOnlyHint=true; mark delete_snippet with destructiveHint=true; mark store_snippet and update_snippet with idempotentHint=true (if they deduplicate by content hash).
• Implement error recovery guidance. When a snippet is not found, return: {error: 'Snippet not found (id=123). Try search_snippets(query=...) to find it by content, or list_recent() to browse recent snippets.'}.