23 tools with consistent verb-noun naming (list_, create_, delete_, describe_, run_, get_, explain_, prepare_, complete_). All tools have descriptions (avg ~80 chars, within baseline 34-392). Input schemas present for all tools with typed parameters and descriptions. However, output schemas are NOT documented, LLMs cannot predict response structure. Error handling is minimal; no recovery guidance or actionable error messages visible. Missing tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk levels. Parameter descriptions are adequate but lack format constraints (e.g., no regex patterns, min/max for numeric fields). No confirmation/dry-run pattern for destructive operations (delete_project, delete_branch).

• Document output schemas for all tools. For list_projects, specify: {projects: [{id, name, org_id, ...}], pagination: {cursor, limit}}. For run_sql, specify: {rows: [...], columns: [...], rowCount: number}.
• Add tool annotations to tool definitions. Mark delete_project and delete_branch with destructiveHint=true. Mark read-only tools (list_projects, describe_project) with readOnlyHint=true. Mark idempotent operations with idempotentHint=true.
• Implement confirmation pattern for destructive operations. Add a dry_run parameter to delete_project and delete_branch, or create a separate confirm_delete_project tool that requires explicit confirmation before executing.
• Formalize parameter constraints in JSON Schema. Add minItems/maxItems for arrays, minimum/maximum for numeric fields (e.g., limit: {type: 'number', minimum: 1, maximum: 400}), and pattern for string fields where applicable.
• Add error recovery guidance to tool descriptions. E.g., 'If the project is not found, call list_projects() to verify the project ID. If the project exists but deletion fails, check for active branches or compute endpoints.'

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

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