Container Security Hardening Skill

A production-focused guide for building, scanning, and running containers securely — from Dockerfile authoring through runtime enforcement and supply chain integrity.

When to Use This Skill

• User mentions Docker security, container hardening, or Dockerfile security review
• User asks about distroless images, non-root containers, or read-only filesystems
• User wants to scan images for CVEs with Trivy, Grype, or Snyk
• User mentions seccomp, AppArmor, Linux capabilities, or runtime security
• User asks "is my Dockerfile secure?" or "how do I reduce my image attack surface?"
• User wants to sign/verify images with Cosign or generate SBOMs
• User asks about Kubernetes pod security, NetworkPolicy, or RBAC hardening
• User says "fix container CVEs" or "harden my container for production"

When NOT to Use This Skill

• The user is primarily asking about GitHub Actions CI/CD → recommend github-actions-advanced
• The user needs general Docker usage help (not security) → recommend docker-expert
• The user is working with Kubernetes orchestration beyond security → recommend kubernetes-architect
• The user needs application-level security (SQL injection, XSS) → recommend api-security-best-practices

Step 1: Understand Context Before Responding

When invoked, first detect the current state:

bash
# Find Dockerfiles in the project
find . -name "Dockerfile*" -not -path "*/node_modules/*" | head -10

# Check for existing security tooling
ls .trivyignore .hadolint.yaml .snyk docker-compose*.yml 2>/dev/null

# Inspect base images currently in use
grep -r "^FROM" $(find . -name "Dockerfile*") 2>/dev/null

# Check if Kubernetes manifests exist
find . -name "*.yaml" -path "*/k8s/*" -o -name "*.yaml" -path "*/manifests/*" | head -10

Then adapt recommendations to:

• The tech stack (Node, Python, Go, Java — affects base image choice)
• Whether this is Docker-only or Kubernetes-deployed
• The CI platform in use (for scanner integration)
• The existing base images and how far they are from best practice

The Five Layers of Container Security

1. Image Build        → Minimal base, no secrets, non-root, read-only FS
2. Image Scanning     → CVE scanning, SBOM, secret detection, Dockerfile lint
3. Runtime Security   → Capabilities, seccomp, AppArmor, resource limits
4. Supply Chain       → Signed images, pinned digests, trusted registries
5. Kubernetes Layer   → Pod Security Admission, NetworkPolicy, RBAC, Kyverno

> Work through layers in order — hardening the image first gives the most leverage. > See references/base-image-comparison.md for a full size/CVE trade-off table.

Layer 1: Dockerfile Hardening

1.1 Use a Minimal Base Image

dockerfile
# ❌ AVOID — massive attack surface (~100–200 CVEs typical)
FROM ubuntu:latest
FROM node:20

# ✅ BETTER — slim variants (glibc, smaller apt footprint)
FROM node:20-slim
FROM python:3.12-slim

# ✅ BEST — distroless (no shell, no package manager, built-in nonroot user)
FROM gcr.io/distroless/nodejs20-debian12
FROM gcr.io/distroless/python3-debian12
FROM gcr.io/distroless/static-debian12   # Go/Rust fully-static binaries

# ✅ ALSO GREAT — Alpine (musl libc; verify app compatibility first)
FROM alpine:3.20

# ✅ ZERO ATTACK SURFACE — for fully static binaries only
FROM scratch

See references/base-image-comparison.md for the full trade-off matrix.

1.2 Multi-Stage Build — Separate Build from Runtime

Never ship build tools, compilers, or dev dependencies in a production image.

dockerfile
# syntax=docker/dockerfile:1

# ── Stage 1: Install & Build ──────────────────────────────
FROM node:20-slim AS builder
WORKDIR /build
COPY package*.json ./
RUN npm ci                          # Install all deps (including devDeps)
COPY . .
RUN npm run build && npm prune --production

# ── Stage 2: Runtime — minimal, no build tools ────────────
FROM gcr.io/distroless/nodejs20-debian12@sha256:<digest>
LABEL org.opencontainers.image.source="https://github.com/org/repo"
LABEL org.opencontainers.image.revision="${BUILD_SHA}"
LABEL org.opencontainers.image.licenses="MIT"
WORKDIR /app
COPY --from=builder --chown=nonroot:nonroot /build/dist        ./dist
COPY --from=builder --chown=nonroot:nonroot /build/node_modules ./node_modules
USER nonroot:nonroot                # UID 65532 — built into distroless
EXPOSE 3000
CMD ["dist/server.js"]

Go / Rust static binary pattern:

dockerfile
FROM golang:1.22-alpine AS builder
WORKDIR /build
COPY go.* ./
RUN go mod download
COPY . .
RUN CGO_ENABLED=0 GOOS=linux go build -ldflags="-s -w" -o app .

FROM scratch                        # Zero attack surface
COPY --from=builder /etc/ssl/certs/ca-certificates.crt /etc/ssl/certs/
COPY --from=builder /build/app /app
USER 65532:65532
ENTRYPOINT ["/app"]

1.3 Run as Non-Root User

dockerfile
# For debian/ubuntu-based images — create dedicated user
RUN groupadd -r appgroup --gid 10001 && \
    useradd -r -g appgroup --uid 10001 --no-log-init appuser

COPY --chown=appuser:appgroup . /app

USER appuser    # Switch before CMD/ENTRYPOINT — never run as root

# ─────────────────────────────────────────────────────────
# For Alpine-based images
RUN addgroup -g 10001 -S appgroup && \
    adduser -u 10001 -S appuser -G appgroup

# For distroless — nonroot (UID 65532) is already built in
USER nonroot:nonroot

1.4 Pin Base Images to Digest

dockerfile
# ❌ UNSAFE — tags are mutable; image can be silently overwritten (supply chain attack)
FROM node:20-slim

# ✅ SAFE — SHA256 digest is cryptographically immutable
FROM node:20-slim@sha256:a1b2c3d4e5f6789abcdef0123456789abcdef0123456789abcdef0123456789ab

Get the current digest:

bash
docker pull node:20-slim
docker inspect node:20-slim --format='{{index .RepoDigests 0}}'

Automate digest pinning with Renovate or Dependabot:

json
// .renovaterc.json
{
  "extends": ["config:base"],
  "dockerfile": { "enabled": true },
  "pinDigests": true
}

1.5 Never Bake Secrets into Images

dockerfile
# ❌ NEVER — secret in ENV or RUN; visible in `docker history` and layer cache
ENV AWS_SECRET_ACCESS_KEY=supersecret
RUN curl -H "Authorization: Bearer $TOKEN" https://api.example.com > config.json
ARG API_KEY                         # Also unsafe — visible in build args history

# ✅ CORRECT — BuildKit secret mount (never persisted in any layer)
# syntax=docker/dockerfile:1
RUN --mount=type=secret,id=api_token \
    curl -H "Authorization: Bearer $(cat /run/secrets/api_token)" \
    https://api.example.com/config > config.json

Build with: docker build --secret id=api_token,src=./token.txt .

Check your image for leaked secrets:

bash
docker history --no-trunc myapp:latest | grep -iE "secret|key|password|token"
trivy image --scanners secret myapp:latest

1.6 Read-Only Filesystem & No New Privileges

dockerfile
# In the Dockerfile — use exec form (no shell interpretation)
ENTRYPOINT ["node", "server.js"]    # ✅ exec form
# ENTRYPOINT /bin/sh -c "node..."  # ❌ shell form — spawns extra process

# Define a HEALTHCHECK
HEALTHCHECK --interval=30s --timeout=5s --start-period=10s --retries=3 \
  CMD ["node", "-e", "require('http').get('http://localhost:3000/health', r => process.exit(r.statusCode === 200 ? 0 : 1))"]

Enforce read-only at runtime (see Layer 3).

1.7 Minimal .dockerignore

dockerignore
# Always exclude these from build context
.git
.github
.env
.env.*
*.pem
*.key
node_modules
__pycache__
.pytest_cache
coverage/
dist/
*.log
.DS_Store
Dockerfile*
docker-compose*
README.md
docs/
tests/

1.8 Full Hardened Dockerfile Example

dockerfile
# syntax=docker/dockerfile:1

# ── Build stage ───────────────────────────────────────────
FROM node:20-slim AS builder
WORKDIR /build
COPY package*.json ./
RUN --mount=type=cache,target=/root/.npm \
    npm ci
COPY . .
RUN npm run build && npm prune --production

# ── Runtime stage ─────────────────────────────────────────
FROM gcr.io/distroless/nodejs20-debian12@sha256:<pin-digest-here>

LABEL org.opencontainers.image.source="https://github.com/org/repo"
LABEL org.opencontainers.image.revision="${BUILD_SHA}"
LABEL org.opencontainers.image.licenses="MIT"

WORKDIR /app
COPY --from=builder --chown=nonroot:nonroot /build/dist        ./dist
COPY --from=builder --chown=nonroot:nonroot /build/node_modules ./node_modules

USER nonroot:nonroot
EXPOSE 3000

HEALTHCHECK --interval=30s --timeout=5s --start-period=10s --retries=3 \
  CMD ["node", "-e", "require('http').get('http://localhost:3000/health', r => process.exit(r.statusCode===200?0:1))"]

CMD ["dist/server.js"]

Layer 2: Image Scanning

2.1 Trivy (Recommended — Fast, Comprehensive)

bash
# Install
brew install trivy                              # macOS
apt install trivy                               # Debian/Ubuntu
tmpdir="$(mktemp -d)"
trap 'rm -rf "$tmpdir"' EXIT
curl -sfL https://raw.githubusercontent.com/aquasecurity/trivy/main/contrib/install.sh \
  -o "$tmpdir/trivy-install.sh"
sed -n '1,160p' "$tmpdir/trivy-install.sh"
sh "$tmpdir/trivy-install.sh"

# Scan an image for CVEs
trivy image myapp:latest

# Fail CI on HIGH/CRITICAL severity
trivy image --exit-code 1 --severity HIGH,CRITICAL myapp:latest

# Scan Dockerfile for misconfigurations
trivy config ./Dockerfile

# Scan entire repo (vulnerabilities + secrets + misconfigs)
trivy fs --scanners vuln,secret,misconfig .

# Generate SBOM (CycloneDX or SPDX)
trivy image --format cyclonedx --output sbom.json myapp:latest
trivy image --format spdx-json  --output sbom.spdx.json myapp:latest

# Ignore specific CVEs (add justification comments)
trivy image --ignorefile .trivyignore myapp:latest

.trivyignore example:

# CVE-2023-1234 — only exploitable via X feature, not used in this app
CVE-2023-1234

# CVE-2023-5678 — fix not yet available; tracked in issue #42
CVE-2023-5678

2.2 Grype (Anchore Alternative)

bash
# Install
tmpdir="$(mktemp -d)"
trap 'rm -rf "$tmpdir"' EXIT
curl -sSfL https://raw.githubusercontent.com/anchore/grype/main/install.sh \
  -o "$tmpdir/grype-install.sh"
sed -n '1,160p' "$tmpdir/grype-install.sh"
sh "$tmpdir/grype-install.sh"

# Scan image
grype myapp:latest

# Fail on critical
grype myapp:latest --fail-on critical

# Output SARIF for GitHub Security tab
grype myapp:latest -o sarif > results.sarif

# Pair with Syft for SBOM generation
syft myapp:latest -o cyclonedx-json > sbom.json
grype sbom:sbom.json                            # Scan the SBOM directly

2.3 Hadolint — Dockerfile Linting

bash
# Run directly
docker run --rm -i hadolint/hadolint < Dockerfile

# With config file
hadolint --config .hadolint.yaml --failure-threshold warning Dockerfile

.hadolint.yaml:

yaml
failure-threshold: warning
ignore:
  - DL3008   # Pin versions in apt-get (allow floating for base layer)
trustedRegistries:
  - gcr.io
  - ghcr.io
  - public.ecr.aws

2.4 Secret Scanning in Images

bash
# Trivy covers secrets too
trivy image --scanners secret myapp:latest

# Dedicated: TruffleHog
trufflehog docker --image myapp:latest

# git-secrets to prevent committing secrets
git secrets --scan

2.5 CI Integration (GitHub Actions — SHA-Pinned)

yaml
permissions:
  contents: read
  security-events: write      # Required for uploading SARIF

jobs:
  security-scan:
    runs-on: ubuntu-24.04
    timeout-minutes: 20
    steps:
      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683  # v4.2.2

      - name: Build image
        run: docker build -t myapp:${{ github.sha }} .

      - name: Lint Dockerfile
        uses: hadolint/hadolint-action@54c9adbab1582c2ef04b2016b760714a4bfde3cf  # v3.1.0
        with:
          dockerfile: Dockerfile
          failure-threshold: warning

      - name: Scan with Trivy
        uses: aquasecurity/trivy-action@6e7b7d1fd3e4fef0c5fa8cce1229c54b2c9bd0d8  # v0.28.0
        with:
          image-ref: myapp:${{ github.sha }}
          format: sarif
          output: trivy-results.sarif
          severity: HIGH,CRITICAL
          exit-code: '1'

      - name: Upload results to GitHub Security tab
        uses: github/codeql-action/upload-sarif@4f3212b61783c3c68e8309a0f18a699764811cda  # v3.27.1
        if: always()          # Upload even if scan found issues
        with:
          sarif_file: trivy-results.sarif

Layer 3: Runtime Security

3.1 docker run Hardening Flags

bash
docker run \
  --read-only \                              # Read-only root filesystem
  --tmpfs /tmp:noexec,nosuid,size=100m \     # Writable tmpfs for /tmp only
  --tmpfs /var/run \                         # For PID files if needed
  --user 10001:10001 \                       # Non-root UID:GID
  --cap-drop ALL \                           # Drop ALL Linux capabilities
  --cap-add NET_BIND_SERVICE \               # Re-add only what's truly needed
  --security-opt no-new-privileges:true \    # Prevent privilege escalation via setuid
  --security-opt seccomp=seccomp.json \      # Custom seccomp profile
  --security-opt apparmor=docker-default \   # AppArmor profile
  --pids-limit 100 \                         # Prevent fork bombs
  --memory 512m \                            # OOM protection
  --memory-swap 512m \                       # Disable swap
  --cpus 1.0 \                               # CPU limit
  --network none \                           # No network (if not needed)
  --health-cmd "curl -f http://localhost:3000/health || exit 1" \
  --health-interval 30s \
  myapp:latest

3.2 Linux Capabilities — What to Drop and Keep

Drop ALL, then explicitly add only what your app requires:

| Capability | Purpose | Keep? | |---|---|---| | NET_BIND_SERVICE | Bind ports < 1024 | Only if binding a privileged port | | CHOWN | Change file ownership | No — set ownership at build time | | SETUID / SETGID | Switch user identity | No — drop always | | SYS_ADMIN | Broad privileged operations | No — most dangerous capability | | NET_ADMIN | Configure network interfaces | No (only network tools) | | SYS_PTRACE | Debug/trace processes | No (only debugger containers) | | DAC_OVERRIDE | Override file permissions | No — runs as correct user | | NET_RAW | Raw sockets (ping) | No (blocked by default seccomp anyway) |

> Most web apps need zero capabilities. --cap-drop ALL alone is often sufficient.

3.3 Docker Compose Hardening

yaml
services:
  app:
    image: myapp:latest
    read_only: true
    user: "10001:10001"
    tmpfs:
      - /tmp:noexec,nosuid,size=100m
      - /var/run:noexec,nosuid,size=10m
    cap_drop:
      - ALL
    cap_add:
      - NET_BIND_SERVICE    # Only if binding port < 1024
    security_opt:
      - no-new-privileges:true
      - seccomp:./references/seccomp-profile-template.json
    pids_limit: 100
    mem_limit: 512m
    memswap_limit: 512m
    cpus: 1.0
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:3000/health"]
      interval: 30s
      timeout: 5s
      retries: 3
      start_period: 10s
    networks:
      - backend
    # Only expose externally if truly required
    # ports: ["8080:8080"]
    restart: unless-stopped
    logging:
      driver: json-file
      options:
        max-size: "10m"
        max-file: "3"

networks:
  backend:
    driver: bridge
    internal: true    # No external connectivity unless needed

3.4 Seccomp Profiles

The Docker default seccomp profile blocks ~44 dangerous syscalls. For stricter control:

bash
# Step 1: Audit syscalls your app actually makes
docker run --security-opt seccomp=unconfined \
  --name audit-run myapp:latest &

# Capture with strace
strace -c -p $(docker inspect --format '{{.State.Pid}}' audit-run)

# Or with sysdig (more container-friendly)
sysdig -p "%syscall.type" container.name=audit-run | sort -u

# Step 2: Build a custom profile from references/seccomp-profile-template.json
# Step 3: Apply it
docker run --security-opt seccomp=references/seccomp-profile-template.json myapp:latest

See references/seccomp-profile-template.json for a minimal starting allowlist for typical web servers.

3.5 AppArmor Profile (Linux hosts)

bash
# Load Docker's default AppArmor profile
sudo apparmor_parser -r /etc/apparmor.d/docker-default

# Apply at runtime
docker run --security-opt apparmor=docker-default myapp:latest

# Generate a custom profile
aa-genprof myapp   # Interactive — run app under aa-complain mode first

Layer 4: Supply Chain Security

4.1 Sign Images with Cosign (Sigstore — Keyless)

bash
# Install cosign
brew install cosign    # macOS
# or: https://github.com/sigstore/cosign/releases

# Sign after push — keyless via OIDC (no long-lived keys)
cosign sign ghcr.io/org/myapp:latest

# Verify before deploy
cosign verify ghcr.io/org/myapp:latest \
  --certificate-identity-regexp="https://github.com/org/repo" \
  --certificate-oidc-issuer="https://token.actions.githubusercontent.com"

GitHub Actions — Sign & Verify Pipeline:

yaml
permissions:
  id-token: write     # Required for OIDC keyless signing
  packages: write

steps:
  - uses: sigstore/cosign-installer@dc72c7d5c4d10cd6bcb8cf6e3fd625a9e5e537da  # v3.7.0

  - name: Sign image (keyless via OIDC)
    run: |
      cosign sign --yes \
        ghcr.io/${{ github.repository }}:${{ github.sha }}
    env:
      COSIGN_EXPERIMENTAL: "true"

  - name: Attach SBOM attestation
    run: |
      cosign attest --yes \
        --predicate sbom.json \
        --type cyclonedx \
        ghcr.io/${{ github.repository }}:${{ github.sha }}

4.2 SBOM Generation & Attestation

bash
# Generate SBOM with Syft
syft myapp:latest -o cyclonedx-json > sbom.json
syft myapp:latest -o spdx-json > sbom.spdx.json

# Attach to image as attestation
cosign attest --predicate sbom.json --type cyclonedx ghcr.io/org/myapp:latest

# Verify SBOM attestation before deployment
cosign verify-attestation \
  --type cyclonedx \
  --certificate-identity-regexp="https://github.com/org/repo" \
  --certificate-oidc-issuer="https://token.actions.githubusercontent.com" \
  ghcr.io/org/myapp:latest

4.3 Use Trusted Registries & Enable Registry Scanning

| Registry | Built-in Scanning | Notes | |---|---|---| | GHCR (GitHub Container Registry) | No (use Trivy in CI) | Best for OSS, OIDC auth | | AWS ECR | Yes (enhanced scanning via Inspector) | Enable per-repo | | GCP Artifact Registry | Yes (Container Analysis) | Enabled by default | | Azure ACR | Yes (Defender for Containers) | Premium tier | | Docker Hub | Yes (limited on free tier) | Avoid for private images |

bash
# Enable ECR enhanced scanning
aws ecr put-registry-scanning-configuration \
  --scan-type ENHANCED \
  --rules '[{"repositoryFilters":[{"filter":"*","filterType":"WILDCARD"}],"scanFrequency":"CONTINUOUS_SCAN"}]'

4.4 Admission Control — Block Unsigned/Unscanned Images

yaml
# Kyverno policy — require signed images before admission
apiVersion: kyverno.io/v1
kind: ClusterPolicy
metadata:
  name: require-signed-images
spec:
  validationFailureAction: Enforce
  rules:
    - name: verify-image-signature
      match:
        resources:
          kinds: [Pod]
      verifyImages:
        - imageReferences:
            - "ghcr.io/org/*"
          attestors:
            - entries:
                - keyless:
                    subject: "https://github.com/org/repo/.github/workflows/*"
                    issuer: "https://token.actions.githubusercontent.com"

Layer 5: Kubernetes Pod Security

> Full reference: references/kubernetes-pod-security.md

5.1 Pod Security Context

yaml
apiVersion: apps/v1
kind: Deployment
metadata:
  name: myapp
  namespace: production
spec:
  replicas: 3
  template:
    spec:
      # ── Pod-level security context ─────────────────────
      securityContext:
        runAsNonRoot: true
        runAsUser: 10001
        runAsGroup: 10001
        fsGroup: 10001
        fsGroupChangePolicy: OnRootMismatch
        seccompProfile:
          type: RuntimeDefault    # Use containerd/runc default seccomp
        supplementalGroups: []

      automountServiceAccountToken: false   # Disable unless needed

      # ── Container-level security context ──────────────
      containers:
        - name: app
          image: ghcr.io/org/myapp@sha256:<digest>   # Always use digest
          securityContext:
            allowPrivilegeEscalation: false
            readOnlyRootFilesystem: true
            capabilities:
              drop: ["ALL"]
              add: []              # Add nothing unless absolutely required
            runAsNonRoot: true
            runAsUser: 10001
            seccompProfile:
              type: RuntimeDefault

          # ── Resource limits (required for restricted PSA) ──
          resources:
            requests:
              memory: "128Mi"
              cpu: "100m"
            limits:
              memory: "512Mi"
              cpu: "500m"

          # ── Writable tmpfs mounts ──────────────────────
          volumeMounts:
            - name: tmp
              mountPath: /tmp
            - name: varrun
              mountPath: /var/run

      volumes:
        - name: tmp
          emptyDir:
            medium: Memory
            sizeLimit: 100Mi
        - name: varrun
          emptyDir:
            medium: Memory
            sizeLimit: 10Mi

5.2 Pod Security Admission (K8s 1.25+)

bash
# Audit existing workloads before enforcing
kubectl label namespace production \
  pod-security.kubernetes.io/audit=restricted \
  pod-security.kubernetes.io/audit-version=latest

# Warn in staging, enforce in production
kubectl label namespace staging \
  pod-security.kubernetes.io/warn=restricted

kubectl label namespace production \
  pod-security.kubernetes.io/enforce=restricted \
  pod-security.kubernetes.io/enforce-version=latest

| PSA Level | What It Blocks | |---|---| | privileged | No restrictions | | baseline | Blocks hostNetwork, hostPID, privileged containers, hostPath | | restricted | Also requires non-root, read-only FS, drops capabilities, seccomp |

5.3 NetworkPolicy — Zero-Trust Networking

yaml
# Step 1: Deny all ingress and egress by default in the namespace
apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  name: default-deny-all
  namespace: production
spec:
  podSelector: {}
  policyTypes: [Ingress, Egress]

---
# Step 2: Selectively allow only required traffic
apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  name: allow-app
  namespace: production
spec:
  podSelector:
    matchLabels:
      app: myapp
  policyTypes: [Ingress, Egress]
  ingress:
    - from:
        - namespaceSelector:
            matchLabels:
              kubernetes.io/metadata.name: ingress-nginx
          podSelector:
            matchLabels:
              app.kubernetes.io/name: ingress-nginx
      ports:
        - port: 3000
  egress:
    - to:
        - podSelector:
            matchLabels:
              app: postgres
      ports:
        - port: 5432
    - to:                 # Allow only cluster DNS
        - namespaceSelector:
            matchLabels:
              kubernetes.io/metadata.name: kube-system
          podSelector:
            matchLabels:
              k8s-app: kube-dns
      ports:
        - port: 53
          protocol: UDP
        - port: 53
          protocol: TCP

5.4 RBAC — Least Privilege

yaml
# Create minimal role — never use wildcards
apiVersion: rbac.authorization.k8s.io/v1
kind: Role
metadata:
  name: app-reader
  namespace: production
rules:
  - apiGroups: [""]
    resources: ["configmaps", "secrets"]
    resourceNames: ["myapp-config"]    # Lock to specific resource names
    verbs: ["get"]                     # Never ["*"]

---
apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
  name: app-reader-binding
  namespace: production
subjects:
  - kind: ServiceAccount
    name: myapp-sa
    namespace: production
roleRef:
  kind: Role
  name: app-reader
  apiGroup: rbac.authorization.k8s.io

bash
# Audit what permissions a service account has
kubectl auth can-i --list --as=system:serviceaccount:production:myapp-sa

# Find overly-permissive cluster roles
kubectl get clusterrolebindings -o json | \
  jq '.items[] | select(.roleRef.name == "cluster-admin") | .subjects'

5.5 Kyverno Policy Examples

yaml
# Require non-root containers
apiVersion: kyverno.io/v1
kind: ClusterPolicy
metadata:
  name: require-non-root
spec:
  validationFailureAction: Enforce
  rules:
    - name: check-run-as-non-root
      match:
        resources:
          kinds: [Pod]
      validate:
        message: "Containers must not run as root (runAsNonRoot: true required)"
        pattern:
          spec:
            containers:
              - securityContext:
                  runAsNonRoot: true

---
# Require image digest pinning
apiVersion: kyverno.io/v1
kind: ClusterPolicy
metadata:
  name: require-image-digest
spec:
  validationFailureAction: Enforce
  rules:
    - name: check-digest
      match:
        resources:
          kinds: [Pod]
      validate:
        message: "Images must be pinned to a SHA256 digest, not just a tag"
        pattern:
          spec:
            containers:
              - image: "*@sha256:*"

---
# Block privileged containers
apiVersion: kyverno.io/v1
kind: ClusterPolicy
metadata:
  name: disallow-privileged
spec:
  validationFailureAction: Enforce
  rules:
    - name: check-privileged
      match:
        resources:
          kinds: [Pod]
      validate:
        message: "Privileged containers are not allowed"
        pattern:
          spec:
            containers:
              - =(securityContext):
                  =(privileged): "false"

Common Pitfalls & Fixes

| Problem | Root Cause | Fix | |---|---|---| | Image runs as root | No USER directive | Add RUN useradd ... and USER appuser | | Secret in docker history | ENV or RUN curl -H "Bearer $TOKEN" | Use RUN --mount=type=secret | | Large image with many CVEs | Full base image (node:20, ubuntu) | Switch to node:20-slim or distroless | | App crashes with --read-only | Writes to /tmp or app directory | Add --tmpfs /tmp for writable temp space | | Trivy scan blocks CI on unfixable CVEs | No ignore file | Add .trivyignore with justified entries | | Container needs SYS_ADMIN | Missing --cap-drop context | Investigate why — almost always avoidable | | Tag-based images drift over time | Mutable tags | Pin to @sha256: digest; use Renovate to update | | K8s pod rejected by PSA | Missing security context fields | Add runAsNonRoot, readOnlyRootFilesystem, allowPrivilegeEscalation: false | | App can't write to filesystem | readOnlyRootFilesystem: true | Mount emptyDir volumes for writable paths |

Security Checklist

Dockerfile

• ] Minimal base image (distroless, slim, or alpine — not full debian/ubuntu)
• ] Multi-stage build — no build tools, devDependencies, or compilers in runtime image
• ] Non-root USER declared before CMD/ENTRYPOINT
• ] Base image pinned to @sha256:... digest (not just tag)
• ] No secrets in ENV, ARG, or RUN commands
• ] HEALTHCHECK defined
• ] OCI labels present (org.opencontainers.image.*)
• ] .dockerignore excludes .git, .env, secrets, tests
• ] ENTRYPOINT uses exec form, not shell form

Image Scanning

• ] Trivy or Grype scan in CI (fails on HIGH/CRITICAL)
• ] Hadolint passes with no warnings
• ] Secret scan run on image (trivy --scanners secret)
• ] SBOM generated and stored
• ] .trivyignore has justified entries for accepted CVEs

Runtime

• ] --read-only filesystem
• ] --cap-drop ALL (add back only what's documented as required)
• ] --security-opt no-new-privileges:true
• ] --security-opt seccomp=<profile> applied
• ] Resource limits set (--memory, --cpus, --pids-limit)
• ] Image signed with Cosign; verified before deploy

Kubernetes

• ] readOnlyRootFilesystem: true
• ] allowPrivilegeEscalation: false
• ] runAsNonRoot: true with explicit UID
• ] capabilities.drop: ["ALL"]
• ] Resource requests and limits defined
• ] automountServiceAccountToken: false
• ] Namespace PSA enforced at restricted level
• ] NetworkPolicy default-deny applied
• ] RBAC uses specific resource names and minimal verbs

Reference Files

• references/base-image-comparison.md — Size, CVE count, shell/pkg-manager trade-offs: distroless vs alpine vs slim vs scratch
• references/seccomp-profile-template.json — Minimal syscall allowlist for typical web servers; start here and extend
• references/kubernetes-pod-security.md — NetworkPolicy, RBAC, OPA/Kyverno policies, service account hardening, PSA

Related Skills

• docker-expert — General Docker usage, Compose orchestration, image optimization
• gha-security-review — Security audit of GitHub Actions workflows
• github-actions-advanced — CI pipeline patterns including scanner integration
• kubernetes-architect — Full Kubernetes architecture, not just security
• api-security-best-practices — Application-level security (injection, auth, OWASP)
• k8s-security-policies — Extended Kubernetes security policies

Limitations

• Use this skill only when the task clearly matches the scope described above.
• Do not treat the output as a substitute for environment-specific penetration testing or a formal security audit.
• Seccomp profiles and AppArmor are Linux-only; macOS/Windows Docker Desktop uses different mechanisms.
• Stop and ask for clarification if required inputs, permissions, safety boundaries, or success criteria are missing.