Server has solid fundamentals: all 5 tools are explicitly registered with schemas, descriptions, and annotations. Tool names follow verb_noun pattern (ssh_*). Descriptions are present and substantive (avg ~150 chars), explaining WHAT each tool does and key constraints. Input schemas use Zod validation with type constraints. However, parameter descriptions are sparse or missing entirely, most params lack actionable guidance on format/range. Output schemas are undocumented (responses are JSON blobs with no declared structure). Error handling returns generic error messages without recovery guidance. Security annotations are present but descriptions don't emphasize risk sufficiently for destructive operations.

• Add detailed parameter descriptions to ssh_upload and ssh_download: 'localPath: Absolute filesystem path (e.g. /home/user/file.txt or C:\Users\file.txt). Must exist for upload; parent directory must exist for download. Symlinks not followed.'
• Document output schemas for all tools. E.g. ssh_connection_info returns {name, host, port, username, authentication, hostVerification, commandTimeoutMs, maxOutputBytes, maxConcurrent}; ssh_exec returns {stdout, stderr, exitCode, timedOut, truncated, signal}.
• Enhance error messages with recovery hints. E.g. 'Command timed out after 5000ms. Try: (1) increase timeoutMs parameter, (2) break command into smaller steps, (3) check if remote host is responsive via ssh_test_connection.'
• Add a dry-run or preview mode to ssh_exec: accept optional 'dryRun: true' parameter that validates syntax without executing, or return a confirmation prompt requiring explicit approval for destructive commands.
• Clarify timeoutMs parameter: 'Timeout in milliseconds (1 - <commandTimeout>). Defaults to server commandTimeout. If exceeded, command is killed and timedOut flag is set. Cannot exceed server limit.'