Skip to content

Security: aws-samples/sample-host-openclaw-on-amazon-bedrock-agentcore

Security

SECURITY.md

Security

Security Architecture

This solution applies defense-in-depth across network, application, identity, and data layers:

  • Network isolation: Private VPC subnets with 7 VPC endpoints; no direct internet exposure for containers
  • Webhook authentication: Cryptographic validation (Telegram secret token, Slack HMAC-SHA256 with replay protection)
  • Per-user microVM isolation: Each user runs in a dedicated Firecracker microVM on AgentCore Runtime
  • STS session-scoped credentials: Container assumes its own role with a session policy restricting S3, DynamoDB, Secrets Manager, and EventBridge to the user's namespace only
  • Secure API key management: Built-in api-keys skill stores user secrets in AWS Secrets Manager (KMS-encrypted, CloudTrail-auditable) — replaces insecure plaintext .env files
  • Encryption: All data encrypted at rest (KMS CMK) and in transit (TLS)
  • Least-privilege IAM: Tightly scoped permissions per component
  • Tool hardening: OpenClaw read tool denied to prevent credential access; exec allowed with STS-scoped blast radius
  • Automated compliance: cdk-nag AwsSolutions checks on every cdk synth

For the complete security architecture — threat model, all 10 defense-in-depth layers, compliance details, operations runbook, and extension roadmap — see docs/security.md.

Reporting Security Issues

See CONTRIBUTING.md for information on reporting security vulnerabilities.

There aren't any published security advisories