Three tools with minimal schema documentation and weak descriptions. All tools lack parameter descriptions despite having inputs. get_target_deployments has empty input schema ({}), making it impossible to validate. kill_random_pod and throttle_cpu_limits have typed parameters but no descriptions explaining what values are valid or how they're used. No output schemas documented. No error handling guidance. Tool descriptions are 10-20 chars below the 50-200 char baseline for LLM optimization. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk profiles.

• Add parameter descriptions to kill_random_pod and throttle_cpu_limits. Example: 'deployment: The name of the target Kubernetes deployment in the default namespace. Must match an existing deployment returned by get_target_deployments().'
• Document output schemas for all three tools. Example for get_target_deployments: 'Returns {"deployments": [{"name": string, "replicas": int, "ready_replicas": int}], "total": int}'
• Add tool annotations: readOnlyHint=true for get_target_deployments, destructiveHint=true for kill_random_pod, idempotentHint=true for throttle_cpu_limits.
• Expand tool descriptions to 100-150 chars with WHEN/WHY context. Example: 'Delete one random running pod from the target deployment to simulate a crash. Use this to test application resilience to pod failures. Requires the deployment name from get_target_deployments().'
• Add error handling guidance. Example: 'If deployment not found, call get_target_deployments() to list available targets. If no pods are running, the operation will fail, check deployment status first.'

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

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