Server exposes 9 ArgoCD tools with basic structure but significant quality gaps. All tools have descriptions (10-50 chars, below the 50-200 char baseline for LLM optimization) and input schemas are present but minimal. Parameter descriptions exist but are generic ('The name of the application'). No output schemas documented. Error handling returns raw error dicts without recovery guidance. Tool names follow verb_noun convention (list_, create_, delete_, etc.) which is correct, but descriptions lack context on WHEN to use each tool or what the LLM should expect. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk levels. Missing pagination for list_applications_tool despite potential for large result sets.

• Add output schemas to all tools. Document the structure of responses (e.g., list_applications returns {applications: [{name, status, syncStatus, ...}], total: int}). This enables LLMs to extract the right fields for downstream calls.
• Expand tool descriptions to 50-200 chars. Include WHEN to use the tool and what it returns. Example: 'List all ArgoCD applications with their sync status. Call this first to discover available apps before operating on a specific one. Returns array of {name, status, syncStatus, lastSyncTime}.'
• Add tool annotations via MCP's readOnlyHint, destructiveHint, and idempotentHint fields. Mark delete_application_tool as destructive, create_application_tool as non-idempotent, and read-only tools accordingly.
• Enhance error messages with recovery guidance. Instead of {'error': 'Not found'}, return {'error': 'Application not found', 'recovery': 'Call list_applications_tool() to see available applications', 'available_apps': ['app1', 'app2']}.
• Add pagination to list_applications_tool and list_event_applications_tool. Accept limit (default 20, max 100) and offset parameters. Return {applications: [...], total: int, hasMore: bool} to prevent context window overflow.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry