Critical gaps across all dimensions. Tool definitions lack visible input schemas in source code, cannot verify parameter types, constraints, or descriptions. Descriptions are present but generic (10-50 chars typical). No evidence of structured output schemas, error handling guidance, or parameter validation. Tool names follow verb_noun convention (positive), but without schemas and detailed descriptions, LLMs cannot reliably select or invoke these tools. The codebase shows tool registration via @mcp.tool() decorator (FastMCP pattern), but actual schema definitions are not visible in provided source. This is a major blocker for production use.

• Add explicit JSON Schema input definitions for all 27 tools. Include type, description, enum constraints, and min/max for every parameter. Example: { 'type': 'object', 'properties': { 'activity_id': { 'type': 'string', 'description': 'Unique activity identifier' } }, 'required': ['activity_id'] }
• Expand tool descriptions to 100-200 characters. Include WHAT the tool does, WHEN to use it (vs similar tools), and any prerequisites. Example: 'Retrieve recent session activity records. Use this to audit agent behavior or debug failed operations. Requires activity_id or time range filter.'
• Document output schemas for all GET/READ tools. Specify field names, types, and pagination (limit, offset, total_count, next_cursor). Example: { 'type': 'object', 'properties': { 'activities': { 'type': 'array', 'items': { 'type': 'object', 'properties': { 'id': { 'type': 'string' }, 'timestamp': { 'type': 'string', 'format': 'date-time' } } } } }
• Add error handling guidance to WRITE tools. Specify which errors are retryable, which require user intervention, and provide recovery hints. Example: 'If disposition already recorded, returns 409 Conflict, this is idempotent; safe to retry.'
• Implement confirmation pattern for destructive operations (record_disposition, clear_handoff, seal_recovery_batch). Add a dry_run parameter or require explicit confirmation_token from a prior preview call.

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

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