Tartarus Harness exposes 15 tools with basic schemas and descriptions, but falls short of production quality. All tools have descriptions (10-80 chars, mostly under 100), but many are terse and lack actionable context. Input schemas are present but minimal: most parameters lack type constraints, enums, or validation guidance. No output schemas documented. Parameter descriptions are sparse or missing entirely. Tools exhibit naming inconsistencies (memory vs memoryId, id vs taskId) that force LLMs to guess field mappings. Error handling is absent, no recovery guidance, no categorization. Security concerns: agent_onboard_mcp accepts arbitrary URLs without validation; no audit trail visible. Composition issues: memory_remember has two alternative fields (content/memory) without clear precedence; plan_update_item has three alternative fields (id/taskId, assignedTo/assigned_to) creating ambiguity. No pagination on search tools despite unbounded result sets. Overall, definitions are skeletal, suitable for internal coordination but not production agent tooling.

• Eliminate alternative parameter names. Choose one canonical field per parameter (e.g., 'id' not 'id'/'taskId'; 'content' not 'content'/'memory'). Document in descriptions why the choice was made.
• Document output schemas for all tools. Specify returned fields, types, and which IDs/references are needed for downstream tool calls. Example: agent_create returns {agent_id, name, profile, created_at}.
• Add pagination to search tools: accept 'limit' (1-100, default 20) and 'offset' or 'cursor'. Return total_count and next_cursor. State the limit in descriptions.
• Add error handling: categorize failures as retryable (network timeout), user-fixable (invalid agent_id), or fatal (permission denied). Return actionable messages: 'Agent not found. Try agent_list() to see available agents.'
• Secure agent_onboard_mcp: validate mcpUrl against a whitelist, require explicit permission scope, log all invocations with caller identity, and implement rate limits (e.g., 5 onboards per agent per hour).