This server has a single tool 'execute_skill_action' with critical definition gaps. The tool description is present but extremely vague (28 chars: 'Execute cloud infrastructure drift remediation actions'). No input schema is visible in the source code, the tool accepts parameters but they are not formally defined with types, constraints, or descriptions. The output is not documented. The implementation is a stub that prints JSON to stdout and does not actually implement MCP protocol semantics. Tool naming is acceptable (verb_noun form), but without schema and parameter documentation, the tool is unusable by agents.

• Define a formal JSON Schema for execute_skill_action inputs. Include 'target_cloud_resource' (string, required, description: 'AWS resource identifier (e.g., aws_security_group.prod_api_gw)'), 'drift_details' (object, optional, description: 'Key-value pairs describing detected drift, e.g., {"unauthorized_port": 22}'), and any other parameters.
• Expand tool description to 100-200 characters. Example: 'Detect and auto-remediate cloud infrastructure drift by comparing live resources against Terraform state. Returns remediation plan and compliance status. Use when infrastructure has drifted from desired state.'
• Document the output schema with typed fields: remediation_plan (string), auto_heal_successful (boolean), security_compliance_status (enum: COMPLIANT_SOC2_TYPE2, NON_COMPLIANT, UNKNOWN).
• Add parameter descriptions in the schema. For 'target_cloud_resource': 'AWS resource identifier in Terraform format (e.g., aws_security_group.prod_api_gw, aws_iam_role.lambda_exec). Required.' For 'drift_details': 'Object describing detected drift. Keys are attribute names, values are unauthorized/unexpected values. Optional; if omitted, tool performs full drift scan.'
• Implement error handling with actionable messages. Example: 'Remediation failed: insufficient IAM permissions. Ensure the execution role has ec2:ModifySecurityGroupRules and iam:UpdateAssumeRolePolicy. Retry after permissions are granted.' Categorize errors as retryable vs. user-fixable.