Critical gaps across all dimensions. Examined src/tools/ directory and catalog.ts: tool definitions lack visible input schemas, parameter descriptions, and output documentation. Sample tools (rsct_status, rsct_load_context, rsct_get_decisions) show no explicit schema registration in source. Tool names use inconsistent prefixes (rsct_*) and lack action verbs (get_*, check_*, request_*, plan_*, phase_*, capture_*, etc. are present but descriptions not visible in provided code). No evidence of parameter type definitions, enums, or constraints. Error handling returns generic JSON errors without recovery guidance. Descriptions appear to be missing or inferred only from tool names.

• Add explicit tool descriptions (50-200 chars) to every tool in catalog.ts. Each must answer: What does it do? When to use it? What does it return? Example: 'rsct_get_decisions: Retrieve all recorded decisions for the current context. Use this to review past choices before making new ones. Returns array of {id, decision, timestamp, rationale}.'
• Define input schemas with JSON Schema for all 40 tools. Every parameter must have type (string|number|boolean|array|object), description, and constraints (enum, minLength, pattern, min/max). Example: {type: 'string', description: 'Phase name (research|spec|code|test|review)', enum: ['research', 'spec', 'code', 'test', 'review']}.
• Document output schemas in tool descriptions or a separate schema registry. Specify which fields are returned, their types, and which are safe to chain to downstream tools. Example: 'Returns {id: string, status: string, timestamp: ISO8601, context_id: string}', the context_id enables chaining to rsct_load_context.
• Improve error handling in CallToolRequestSchema handler. Replace generic errors with actionable guidance. Example: 'Tool rsct_get_decisions requires context_id parameter. Call rsct_load_context first to establish context, then retry with context_id.'
• Rename tools for clarity. Replace 'rsct_status' with 'rsct_get_status' or 'rsct_check_status'. Replace 'rsct_load_context' with 'rsct_set_context' or 'rsct_initialize_context'. Verb-first naming helps LLMs parse intent from the name alone.

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

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