VaultHash has clear, well-structured tool definitions with proper Zod schemas and consistent error handling. All 5 tools have descriptions (10 - 50 chars) and input schemas with typed parameters. However, descriptions are terse (below the 50 - 200 char LLM-optimized baseline), parameter descriptions lack context about constraints and use cases, and output schemas are not formally documented. Tools follow verb_noun naming (file_hash, verify_hash, etc.) but descriptions do not explain WHEN to use each tool or dependencies between them. No tool annotations (readOnlyHint, idempotentHint) despite all being read-only. Error handling wraps exceptions but does not guide recovery or categorize errors.

• Expand tool descriptions to 50 - 150 chars. Example: 'file_hash: Compute a cryptographic hash (sha256, sha1, sha512, or md5) of a local file. Use this to verify file integrity, detect duplicates, or create checksums for archival. Returns the hash in standard format (algo hexdigest).'
• Add parameter descriptions for 'algo': 'Hash algorithm. sha256 (default, recommended for security), sha1 (legacy), sha512 (stronger), md5 (weak, legacy only). Most use cases should use sha256.'
• Document output schemas in tool descriptions. Example: 'Returns a string in format "sha256 <40-char-hex>" for file_hash, or "MATCH ✅" / "MISMATCH ❌" for verify_hash.'
• Add readOnlyHint: true to all tool definitions to signal to agents that these tools are safe to call without side effects.
• Improve error messages with recovery guidance. Example: catch (e) { if (e.code === 'ENOENT') return text('File not found. Check the path and try again.'); else return { isError: true, content: [{ type: 'text', text: 'Error reading file: ' + e.message }] }; }