29 tools with consistent verb-noun naming (search, create, update, delete, list). All tools have descriptions (avg ~60 chars, baseline 194). Input schemas present with type definitions and parameter descriptions. However, descriptions are significantly shorter than production baseline (194 chars), most are terse (20-40 chars), limiting LLM context for tool selection. Output schemas not documented. Error handling present but generic. Missing idempotency guidance for write operations. Destructive operations (delete-user, delete-group, delete-entry, reset-suffix) require 'confirm' parameter but lack dry-run alternatives.

• Expand tool descriptions to 100-200 chars. Include WHAT the tool does, WHEN to use it (vs similar tools), and key output fields. E.g., 'Get a specific LDAP entry by DN. Returns all attributes unless filtered. Use after search() to fetch full details. Returns: dn, attributes object, modifyTimestamp.'
• Document output schemas for all tools. Specify returned fields, types, and pagination info. E.g., search() returns {entries: [{dn, attributes}], total, nextCursor}. This enables LLM planning and chaining.
• Replace 'confirm' string parameter with explicit confirmation flow. Use MRTR (Multi Round-Trip Requests) with result 'input_required' to ask user before destructive ops. Or add a dry_run boolean parameter that returns what would be deleted without executing.
• Add idempotency guidance to write operations. Document which operations are idempotent (safe to retry) and which are not. For non-idempotent ops, provide a revision/etag field for optimistic locking (already present for some tools, extend to all).
• Move password handling to server-side secret injection. Remove 'password' from create-user and set-password parameters. Instead, accept a 'password_secret_name' or use environment variables. Document in server config how to inject credentials securely.