InstaServer has 12 well-named tools with clear verb-noun patterns (deploy_pod, create_pod, delete_pod, etc.). All tools have descriptions (avg ~150 chars, within baseline 194). Input schemas are present and use Zod with type constraints (enums for presets/actions, string for names). However, several parameter descriptions are minimal or missing entirely. Output schemas are not documented, LLMs cannot predict response structure. Error handling returns generic fail() responses without recovery guidance. No tool annotations (readOnlyHint/destructiveHint) despite clear risk levels. The deploy_pod tool is well-designed for composition, but exec_command and write_file lack safety guardrails.

• Document output schemas for all tools. Example: deploy_pod returns {public: string, status: string, pod_id: string, created_at: ISO8601}. This enables agents to chain calls and extract required fields.
• Add tool annotations to all tools. Mark list_presets, list_pods, get_pod, get_logs, list_files, read_file as readOnlyHint=true. Mark delete_pod, manage_pod(stop/restart), deploy_pod as destructiveHint=true. Mark create_pod, deploy_pod as idempotentHint=true where applicable.
• Enhance error responses with recovery guidance. Instead of fail(err), return {isError: true, message: '...', recovery: 'Try list_pods() to verify pod exists', retryable: true/false}. Categorize errors as user-fixable, retryable, or fatal.
• Add input validation and security notes to exec_command and write_file descriptions. Example: 'Command is executed in a sandboxed container. Paths are restricted to the app root; traversal attempts are blocked. Commands timeout after 30s.'
• Expand parameter descriptions with constraints and examples. Example: name parameter → 'DNS-safe pod name (2-63 chars, lowercase alphanumeric + hyphens, no leading/trailing hyphens). Example: my-app-v1'. lines parameter → 'Number of log lines (1-5000, default 100)'.