# Iac Container Security

> Audit infrastructure-as-code and container security including Terraform/OpenTofu/Pulumi configurations, Dockerfile hardening, Kubernetes manifests, base image hygiene, container scanning, secrets in IaC, IAM policies, network exposure, and runtime security context. Multi-cloud (AWS, GCP, Azure). Use this skill whenever the user asks about Terraform security, tfsec, Checkov, Trivy, Dockerfile hard…

- **Type:** Skill
- **Install:** `agentstack add skill-hlsitechio-claude-skills-security-iac-container-security`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [hlsitechio](https://agentstack.voostack.com/s/hlsitechio)
- **Installs:** 0
- **Category:** [Cloud & Infrastructure](https://agentstack.voostack.com/c/cloud-infrastructure)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [hlsitechio](https://github.com/hlsitechio)
- **Source:** https://github.com/hlsitechio/claude-skills-security/tree/main/saas-security-pack/iac-container-security

## Install

```sh
agentstack add skill-hlsitechio-claude-skills-security-iac-container-security
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# IaC and Container Security Audit

Audit infrastructure-as-code repositories and container images for security issues that compromise deployments. Defensive find-and-fix.

## When this skill applies

- Reviewing Terraform / OpenTofu / Pulumi / CloudFormation / ARM templates
- Hardening Dockerfile or reviewing built container images
- Reviewing Kubernetes manifests (Deployments, Services, NetworkPolicies, RBAC)
- Auditing IAM policies for least-privilege violations
- Interpreting output from tfsec, Checkov, Trivy, kubesec, kube-bench
- Reviewing cloud configuration drift (actual cloud state vs IaC)

Use other skills for: workflow/CI security (`github-supply-chain`), app-layer concerns (`saas-code-security-review`, `saas-api-security`).

## Workflow

Follow `../_shared/audit-workflow.md`. IaC-specific notes below.

### Phase 1: Scope confirmation

- Which IaC tools (Terraform, Pulumi, CDK, CloudFormation, Bicep, Helm)?
- Which container runtime (Docker, containerd, ECS, k8s, Cloud Run, App Runner)?
- Which cloud(s)?
- Is the IaC the source of truth, or is config drift expected?

### Phase 2: Inventory

```bash
# IaC files
find . -name '*.tf' -o -name '*.tfvars' \
  -o -name 'Pulumi.yaml' -o -name '*.bicep' \
  -o -name 'cloudformation*.yml' -o -name 'cloudformation*.yaml' \
  -o -name 'Dockerfile*' -o -name 'docker-compose*.yml' \
  -o -path '*/k8s/*.yaml' -o -path '*/manifests/*.yaml' \
  | grep -v node_modules

# Run security scanners (read-only)
trivy fs --severity HIGH,CRITICAL .
checkov -d . --quiet --output cli
tfsec . --soft-fail
docker scout cves --only-severity high,critical local://your-image:tag
```

### Phase 3: Detection — the checks

#### Dockerfile — see `references/dockerfile-hardening.md`

- **IACS-DF-1** Pinned base image (`FROM node:20.11.1-alpine` not `FROM node:latest`).
- **IACS-DF-2** Non-root user (`USER appuser`) declared and used at runtime.
- **IACS-DF-3** No secrets baked into image layers (`ENV` with credential, `COPY .env`).
- **IACS-DF-4** No unnecessary tools (curl, wget, git, build tools) in production stage.
- **IACS-DF-5** Multi-stage build separating build environment from runtime.
- **IACS-DF-6** `HEALTHCHECK` defined.
- **IACS-DF-7** `.dockerignore` excluding `.git`, `.env`, `node_modules` (avoids bloat AND secret leakage).
- **IACS-DF-8** Distroless or minimal base image (`gcr.io/distroless/`, `alpine`, `chainguard`).
- **IACS-DF-9** No `ADD` for URLs (use `RUN curl` with verification, or better, no fetches at all).
- **IACS-DF-10** Image scanned for CVEs in CI; build fails on High/Critical without exception.

#### Kubernetes manifests — see `references/k8s-hardening.md`

- **IACS-K8S-1** `runAsNonRoot: true` and `runAsUser: ` in pod/container securityContext.
- **IACS-K8S-2** `readOnlyRootFilesystem: true` (with emptyDir for writable paths if needed).
- **IACS-K8S-3** `allowPrivilegeEscalation: false`.
- **IACS-K8S-4** `capabilities: { drop: [ALL] }` and add only what's needed.
- **IACS-K8S-5** No `privileged: true` (or documented justification).
- **IACS-K8S-6** `hostNetwork`, `hostPID`, `hostIPC` all false.
- **IACS-K8S-7** Resource requests AND limits set on every container.
- **IACS-K8S-8** NetworkPolicy default-deny in every namespace; specific allow rules.
- **IACS-K8S-9** ServiceAccount per workload (not shared `default` SA).
- **IACS-K8S-10** Pod Security Admission level `restricted` for application namespaces.
- **IACS-K8S-11** No secrets in `env` from literal values; use `valueFrom: secretKeyRef`.
- **IACS-K8S-12** Image pull from a private registry with image digest pinning (`image: registry/foo@sha256:...`).

#### Terraform / OpenTofu

- **IACS-TF-1** State stored in remote backend with encryption + state locking (S3 + DynamoDB, GCS, Azure storage).
- **IACS-TF-2** State backend access restricted to deployer role/team.
- **IACS-TF-3** No secrets in `.tf` or `.tfvars` committed to git. Use Vault / SOPS / cloud secret manager.
- **IACS-TF-4** Provider versions pinned in `required_providers`.
- **IACS-TF-5** Modules sourced from trusted registries with version pins.
- **IACS-TF-6** `terraform plan` reviewed in CI before apply; manual approval for production apply.
- **IACS-TF-7** Drift detection runs (e.g., scheduled `terraform plan` alert).

#### Common cloud resource issues

**Storage (S3/GCS/Azure Blob):**
- **IACS-STOR-1** No `Public Access = ON` without explicit business need.
- **IACS-STOR-2** Bucket policies don't allow `Principal: "*"` for sensitive buckets.
- **IACS-STOR-3** Server-side encryption enabled.
- **IACS-STOR-4** Versioning enabled on critical buckets.
- **IACS-STOR-5** Lifecycle rules archive/delete old data.
- **IACS-STOR-6** Logging to a separate (audit) bucket.

**Compute (EC2/GCE/Azure VM):**
- **IACS-COMP-1** No SSH 0.0.0.0/0 in security group / firewall.
- **IACS-COMP-2** IMDSv2 enforced (AWS).
- **IACS-COMP-3** Instance role scoped, not full admin.
- **IACS-COMP-4** Disks encrypted.
- **IACS-COMP-5** Public IP only when required.

**Networking:**
- **IACS-NET-1** Default VPC/subnets not used for production.
- **IACS-NET-2** Security groups / firewall rules scoped (no `0.0.0.0/0` on internal services).
- **IACS-NET-3** Subnets logically separated (public/private/data).
- **IACS-NET-4** Egress controlled (especially for compute that processes user data).
- **IACS-NET-5** VPC flow logs enabled.

**Databases (RDS/CloudSQL/Azure DB):**
- **IACS-DB-1** Not publicly accessible.
- **IACS-DB-2** Encryption at rest and in transit.
- **IACS-DB-3** Backup retention ≥ 7 days.
- **IACS-DB-4** Multi-AZ / replica for production.
- **IACS-DB-5** Master password from secret manager.
- **IACS-DB-6** Slow query / audit logging where supported.

**IAM:**
- **IACS-IAM-1** No `*:*` policies on production roles.
- **IACS-IAM-2** No `Principal: "*"` on resource policies without explicit `aws:SourceArn`/equivalent constraint.
- **IACS-IAM-3** Roles per-workload, not shared.
- **IACS-IAM-4** No long-lived access keys (use roles/OIDC).
- **IACS-IAM-5** MFA required for sensitive operations.
- **IACS-IAM-6** Permission boundaries on roles delegated to applications.

**Secrets / KMS:**
- **IACS-SEC-1** All secrets in a managed secret store (Secrets Manager, GCP Secret Manager, Azure Key Vault, Vault).
- **IACS-SEC-2** KMS keys with restrictive key policies; rotation enabled.
- **IACS-SEC-3** Secret retrieval logged.
- **IACS-SEC-4** No hard-coded ARNs to other accounts that grant cross-account access without audit.

**Logging and monitoring:**
- **IACS-LOG-1** CloudTrail / Cloud Audit Logs / Activity Log enabled and exported to a separate account/project.
- **IACS-LOG-2** Log integrity validation enabled.
- **IACS-LOG-3** GuardDuty / Security Command Center / Defender for Cloud enabled.
- **IACS-LOG-4** Alerts on root login, IAM policy changes, security group changes.

### Phase 4: Triage

Critical class examples:
- Public S3 bucket containing user data
- IAM role with `*:*` policy attached to a production workload
- K8s pod with `privileged: true` reachable from public ingress
- Database publicly accessible
- IMDSv1 enabled with sensitive instance role

### Phase 5: Report

Use `../_shared/findings-schema.md`. Prefix IDs with `IACS-`.

## Tool integration

For each scanner the audit uses, document the command, the config file, and any baseline suppressions:

```yaml
# .checkov.yaml
skip-check:
  - CKV_AWS_18   # S3 access logging — covered by central log aggregation
```

Suppressions should have inline justifications. Treat unjustified suppressions as findings.

## References

- `references/dockerfile-hardening.md` — Multi-stage, non-root, distroless, scanning
- `references/k8s-hardening.md` — Pod Security Standards, network policies, RBAC
- `references/iam-least-privilege.md` — Policy design patterns, multi-cloud

## Scripts

- `scripts/run_iac_scanners.sh` — Runs tfsec, Checkov, Trivy in sequence with consistent output

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [hlsitechio](https://github.com/hlsitechio)
- **Source:** [hlsitechio/claude-skills-security](https://github.com/hlsitechio/claude-skills-security)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** yes
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** yes
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-hlsitechio-claude-skills-security-iac-container-security
- Seller: https://agentstack.voostack.com/s/hlsitechio
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.