CoWorker MCP exposes 8 tools with basic schemas but significant quality gaps. Tool names lack clear action verbs (skill_card_query vs search_skills, context_sync vs synchronize_context). Most parameter descriptions are minimal (10-30 chars), below the 72-char baseline. Output schemas are undocumented, callers cannot predict response structure. Error handling is absent from visible code. The skill_to_mcp_tool() conversion infers types from string names ('int', 'float', 'bool') rather than reading explicit schema definitions, suggesting incomplete parameter metadata. No tool has documented return types or pagination guidance. Security concerns: session_id and task_id parameters suggest stateful operations, but no audit trail or permission checks are visible.

• Rename tools to verb_noun pattern: skill_card_query → search_skills, skill_card_get → get_skill, context_sync → synchronize_context, context_query → query_context. This enables LLMs to infer intent from names alone.
• Expand parameter descriptions to 50-150 chars. For each param, state: what it controls, valid format/values, and when it's required. E.g., 'tags_filter: Array of skill category tags (e.g., ["data-processing", "api-integration"]). Leave empty to return all skills.'
• Document output schemas for all tools. Specify field names, types, and whether results are paginated. E.g., skill_query returns {skills: [{name, description, tags, input_schema}], total_count}. Include chaining IDs (e.g., task_request returns task_id for polling).
• Add error handling documentation. Define error codes (e.g., SKILL_NOT_FOUND, TIMEOUT, PERMISSION_DENIED), recovery guidance, and retryability. E.g., 'TIMEOUT (retryable): Skill execution exceeded timeout_seconds. Retry with increased timeout or break task into smaller steps.'
• Declare permissions and audit requirements. Add scope descriptions (e.g., 'read:skills', 'write:tasks') and log all calls with user/agent ID, tool name, parameters, timestamp, and result. This enables least-privilege configuration and compliance.

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

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