Spawnr has 8 tools with complete input schemas and descriptions. Naming follows verb_noun convention (search, show, check, hire, whoami, login, logout, fund). Descriptions are present and contextual (avg ~100 chars), meeting baseline minimums. However, several tools lack output schema documentation, parameter descriptions are sometimes sparse, and error handling guidance is minimal. The 'hire' tool combines multiple concerns (detection + installation + config writing), violating single-responsibility. No tool annotations (readOnlyHint/destructiveHint) despite clear risk levels. Parameter constraints are present but inconsistently documented.

• Document output schemas for all tools. E.g., search returns {query: string, agents: [{chain, id, name, tier, score, url}]}. This enables LLMs to extract fields reliably.
• Split 'hire' into three tools: detect_ide (returns detected IDEs), write_mcp_config (writes config to one IDE), link_account (pairs install with thespawn.io account). Each has one responsibility.
• Add tool annotations: mark hire/login/logout/fund as destructive=true; mark search/show/check/whoami/fund as readOnly=true. Use idempotent=true for login (re-linking is safe).
• Add enum constraints to parameters: 'only' in hire should be enum [claude, cursor, windsurf, codex, openclaw]; 'chain' in whoami should list valid chains (base, arbitrum, bsc, celo, tempo).
• Enhance error messages with recovery paths. E.g., 'Agent not found on base. Try: (1) search for similar agents, (2) check the chain slug, or (3) verify the agent ID is correct.'