God Clause exposes 14 tools with significant definition gaps. Most tools lack proper input schemas, parameters are documented in descriptions but not in structured JSON Schema format. Tool names are reasonable (health, ready, evaluate, enforce, audit, metrics, validate, lint, summarize, diff, evaluate_cli, audit_verify, audit_export) but descriptions are inconsistent in depth. The evaluate and enforce tools have detailed parameter descriptions, but many others (health, ready, contracts, metrics) have minimal or missing schema definitions. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) are present. Error handling guidance is absent, tools do not explain recovery paths or categorize errors. Output schemas are not documented, forcing LLMs to infer response structure. The audit tool's filtering parameters are well-described, but most tools lack the depth needed for confident LLM selection and invocation.

• Add explicit JSON Schema definitions for all tool inputs. For health and ready (empty inputs), document as {"type": "object", "properties": {}, "required": []}. For audit, formalize limit (number, 1-1000), offset (number, >=0), allowed (boolean), action (string enum), user_id (string), contract_name (string), start_time (ISO 8601 string), end_time (ISO 8601 string).
• Add tool annotations: mark health, ready, contracts, audit, metrics, validate, lint, summarize, diff, evaluate_cli, audit_verify as readOnlyHint=true. Mark enforce as destructiveHint=true. Mark evaluate as idempotentHint=true.
• Document output schemas for all tools. E.g., audit returns {"entries": [{"id", "timestamp", "action", "allowed", "user_id", "contract_name"}], "total": number, "limit": number, "offset": number}. evaluate returns {"allowed": boolean, "reason": string, "tags_matched": [string], "evaluation_time_ms": number}.
• Add error handling guidance to enforce, validate, lint. E.g., 'If PolicyViolationError is raised, the action was denied. Review the contract rules and caller permissions, then retry with modified input or escalate to a human reviewer.'
• Constrain file path parameters. Add to validate, lint, summarize, diff, audit_verify, audit_export: 'file must be a relative path within the contracts or audit directory (no ../ traversal). Supported formats: .yaml, .yml, .json for contracts; .jsonl for audit files.'

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

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