WAF-AGN-030 – Reasoning Transparency

Säule: Agentic | Schweregrad: Mittel | Kategorie: Logging Monitoring | Automatisierbar: High

Beschreibung

Agent decisions must be traceable – not a black box. Every decision step MUST be logged including chain of thought, reasoning, and decision outcome. Audit trails MUST preserve the full context of agent decision-making for compliance and debugging.

Begründung

Risiko	Beschreibung
Risk	Opaque agent decisions prevent root-cause analysis and regulatory audits.
Risk	Missing reasoning chains make it impossible to demonstrate compliance with business rules.

Risiko

Beschreibung

Risk

Opaque agent decisions prevent root-cause analysis and regulatory audits.

Risk

Missing reasoning chains make it impossible to demonstrate compliance with business rules.

Anforderungen

Agent decisions must be traceable – not a black box.
Every decision step MUST be logged including chain of thought, reasoning, and decision outcome.
Audit trails MUST preserve the full context of agent decision-making for compliance and debugging.

Implementierungsleitfaden

Document the requirement in a version-controlled policy or standard.
Assign ownership and establish a review cadence (e.g., quarterly).
Implement the required configuration as code in Terraform or equivalent IaC.

Reifegrade

Level	Name	Kriterien
L1	Ad hoc – control not consistently applied	* Manual evidence exists for some workloads, but coverage is unknown.
L2	Standardised – policy defined and communicated	* A version-controlled policy, ADR, or standard describes the requirement.
L3	Integrated – automated validation in CI/CD	* Terraform checks run automatically on pull requests and report results.
L4	Auditable – metrics and exceptions tracked	* Control coverage, failures, and approved exceptions are reported quarterly.
L5	Optimised – continuous improvement based on data	* Control effectiveness is measured and iteratively improved.

Level

Name

Kriterien

Ad hoc – control not consistently applied

* Manual evidence exists for some workloads, but coverage is unknown.

Standardised – policy defined and communicated

* A version-controlled policy, ADR, or standard describes the requirement.

Integrated – automated validation in CI/CD

* Terraform checks run automatically on pull requests and report results.

Auditable – metrics and exceptions tracked

* Control coverage, failures, and approved exceptions are reported quarterly.

Optimised – continuous improvement based on data

* Control effectiveness is measured and iteratively improved.

Checks

waf-agn-030.tf.aws.bedrock-agent-logging-enabled

Engine: terraform | Provider: aws | Automated: Ja

AWS Bedrock agent reasoning must be logged

AWS Bedrock agents must have inference logging enabled to capture chain of thought, reasoning steps, and decision outcomes. Logs must include prompt, model response, and metadata.

Scope: resource — aws_bedrockagent, aws_cloudwatch_log_group

Assertions: * name not empty — Agent reasoning logging must be configured.

Remediation: Enable inference logging in AWS Bedrock console or via API. Configure CloudWatch log group for agent inference logs. Set log level to INCLUDE_INFERENCES for full chain of thought.

Compliant

resource "aws_cloudwatch_log_group" "agent_reasoning" {
  name = "/aws/bedrock/inference-logs/reasoning"
}

resource "aws_bedrockagent" "reasoning_agent" {
  agent_name = "reasoning-enabled-agent"
  # Logging enabled via CloudWatch configuration
}

Non-Compliant

resource "aws_bedrockagent" "no_reasoning_log" {
  agent_name = "black-box-agent"
  # No reasoning logging configured
  # WAF-AGN-030 Violation
}

waf-agn-030.tf.azurerm.diagnostic-setting-enabled

Engine: terraform | Provider: azurerm | Automated: Ja

Azure AI services must have diagnostic settings enabled for traceability

Azure Cognitive Services / AI services accounts must have diagnostic settings enabled to capture audit logs including requests, responses, and token usage.

Scope: resource — azurerm_monitor_diagnostic_setting

Assertions: * target_resource_id not empty — Diagnostic setting must target an AI services account. * log_analytics_workspace_id not empty — Diagnostic setting must forward logs to a Log Analytics workspace. * enabled_log attribute exists — Diagnostic setting must enable audit logs.

Remediation: Add azurerm_monitor_diagnostic_setting for the azurerm_cognitive_account to forward audit logs to a Log Analytics workspace.

Compliant

resource "azurerm_log_analytics_workspace" "ai_logs" {
  name                = "ai-logs"
  location            = azurerm_resource_group.main.location
  resource_group_name = azurerm_resource_group.main.name
  sku                 = "PerGB2018"
  retention_in_days   = 90
}

resource "azurerm_monitor_diagnostic_setting" "ai_audit" {
  name                       = "ai-audit-logs"
  target_resource_id         = azurerm_cognitive_account.ai.id
  log_analytics_workspace_id = azurerm_log_analytics_workspace.ai_logs.id

  enabled_log {
    category = "Audit"
  }

  metric {
    category = "AllMetrics"
  }
}

Non-Compliant

resource "azurerm_cognitive_account" "ai" {
  name = "ai-services-account"
  # No diagnostic setting configured
  # WAF-AGN-030 Violation: agent decisions are not traceable
}

waf-agn-030.tf.google.vertex-agent-logging-enabled

Engine: terraform | Provider: google | Automated: Ja

Google Vertex AI agent reasoning must be logged

Google Vertex AI agents must have debugging and reasoning logs enabled. Logs must capture chain of thought, tool calls, and decision paths.

Scope: resource — google_vertex_ai_agent, google_logging_project_sink

Assertions: * name not empty — Agent reasoning logging must be configured.

Remediation: Enable Cloud Logging for Vertex AI. Configure log include patterns to capture agent reasoning. Set log level to DEBUG for full visibility.

Compliant

resource "google_logging_project_sink" "vertex_reasoning" {
  name = "vertex-reasoning-logs"
  destination = "logging.googleapis.com/projects/${data.google_project.project.id}/logs/vertex-ai-reasoning"
  filter = "resource.type=vertex_ai.custom_model OR resource.type=vertex_ai.endpoint"
}

resource "google_vertex_ai_agent" "logging_agent" {
  display_name = "reasoning-logged"
}

Non-Compliant

resource "google_vertex_ai_agent" "no_logging" {
  display_name = "black-box"
  # No reasoning logging
  # WAF-AGN-030 Violation
}

waf-agn-030.tf.aws.tracing-enabled

Engine: terraform | Provider: aws | Automated: Ja

Distributed tracing must be enabled for agents

Agent operations must be traced using AWS X-Ray or similar distributed tracing. Traces must include agent ID, operation type, latency, and decision points.

Scope: resource — aws_lambda_function, aws_bedrockagent

Assertions: * environment.variables.AWS_XRAY_TRACING_NAME not empty — Distributed tracing must be enabled via X-Ray.

Remediation: Enable X-Ray tracing for Bedrock agent invocations. Use AWS SDK with tracing enabled (AWS_XRAY_TRACING_NAME environment variable). Instrument agent code with X-Ray SDK.

Compliant

resource "aws_lambda_function" "agent_handler" {
  function_name = "agent-handler"
  # Tracing enabled via environment variable
  environment {
    variables = {
      AWS_XRAY_TRACING_NAME = "agent-service"
    }
  }
}

Non-Compliant

resource "aws_lambda_function" "untraced" {
  function_name = "untraced-agent"
  # No X-Ray tracing enabled
  # WAF-AGN-030 Violation
}

waf-agn-030.tf.stackit.agentic-decisions-traceable

Engine: terraform | Provider: stackit | Automated: Ja

Agent decisions must be traceable – not a black box for stackit

Agent decisions must be traceable – not a black box must be documented in IaC via a agentic-decisions-traceable tag/label on stackit_project. This validates the control is tracked as code.

Scope: resource — stackit_project

Assertions: * labels key exists — stackit_project is missing the agentic-decisions-traceable tag/label.

Remediation: Add a agentic-decisions-traceable tag/label to the stackit_project resource to show the stackit implementation of WAF-AGN-030 is tracked in IaC.

Compliant

resource "stackit_project" "example" {
  labels = {
    agentic-decisions-traceable = "true"
  }
}

Non-Compliant

resource "stackit_project" "example" {
  # Missing agentic-decisions-traceable labels
}

waf-agn-030.tf.ovh.agentic-decisions-traceable

Engine: terraform | Provider: ovh | Automated: Ja

Agent decisions must be traceable – not a black box for ovh

Agent decisions must be traceable – not a black box must be documented in IaC via a agentic-decisions-traceable tag/label on ovh_cloud_project. This validates the control is tracked as code.

Scope: resource — ovh_cloud_project

Assertions: * tags key exists — ovh_cloud_project is missing the agentic-decisions-traceable tag/label.

Remediation: Add a agentic-decisions-traceable tag/label to the ovh_cloud_project resource to show the ovh implementation of WAF-AGN-030 is tracked in IaC.

Compliant

resource "ovh_cloud_project" "example" {
  tags = {
    agentic-decisions-traceable = "true"
  }
}

Non-Compliant

resource "ovh_cloud_project" "example" {
  # Missing agentic-decisions-traceable tags
}

waf-agn-030.tf.hcloud.agentic-decisions-traceable

Engine: terraform | Provider: hcloud | Automated: Ja

Agent decisions must be traceable – not a black box for hcloud

Agent decisions must be traceable – not a black box must be documented in IaC via a agentic-decisions-traceable tag/label on hcloud_server. This validates the control is tracked as code.

Scope: resource — hcloud_server

Assertions: * labels key exists — hcloud_server is missing the agentic-decisions-traceable tag/label.

Remediation: Add a agentic-decisions-traceable tag/label to the hcloud_server resource to show the hcloud implementation of WAF-AGN-030 is tracked in IaC.

Compliant

resource "hcloud_server" "example" {
  labels = {
    agentic-decisions-traceable = "true"
  }
}

Non-Compliant

resource "hcloud_server" "example" {
  # Missing agentic-decisions-traceable labels
}

Evidenz

Art Erforderlich Beschreibung

Art	Erforderlich	Beschreibung
`iac`	Erforderlich	Terraform configuration implementing the control.
`governance`	Erforderlich	Policy, ADR, or standard documenting the requirement.
`config`	Optional	Cloud console or API output confirming runtime state.

iac

Erforderlich

Terraform configuration implementing the control.

governance

Erforderlich

Policy, ADR, or standard documenting the requirement.

config

Optional

Cloud console or API output confirming runtime state.

WAF-AGN-030 – Reasoning Transparency

Beschreibung

Begründung

Anforderungen

Implementierungsleitfaden

Reifegrade

Checks

waf-agn-030.tf.aws.bedrock-agent-logging-enabled

waf-agn-030.tf.azurerm.diagnostic-setting-enabled

waf-agn-030.tf.google.vertex-agent-logging-enabled

waf-agn-030.tf.aws.tracing-enabled

waf-agn-030.tf.stackit.agentic-decisions-traceable

waf-agn-030.tf.ovh.agentic-decisions-traceable

waf-agn-030.tf.hcloud.agentic-decisions-traceable

Evidenz

Verwandte Controls