Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Investigate compromised Docker containers by analyzing images, layers, volumes, logs, and runtime artifacts to identify malicious activity and evidence.
.claude/skills/analyzing-docker-container-forensics/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | — | — |
| case-20 | ✗→✓ | ▲ Improved | — | — |
| case-19 | ✓→✓ | = Same ✓ | — | — |
| case-21 | ✗→✗ | = Same ✗ | — | — |
| case-08 | ✗→✗ | = Same ✗ | — | — |
bash# List all containers (including stopped) docker ps -a --no-trunc > /cases/case-2024-001/docker/container_list.txt # Inspect the compromised container CONTAINER_ID="abc123def456" docker inspect $CONTAINER_ID > /cases/case-2024-001/docker/container_inspect.json # Export container filesystem as tarball (preserves current state) docker export $CONTAINER_ID > /cases/case-2024-001/docker/container_export.tar # Create an image from the container's current state docker commit $CONTAINER_ID forensic-evidence:case-2024-001 docker save forensic-evidence:case-2024-001 > /cases/case-2024-001/docker/container_image.tar # Capture container logs docker logs $CONTAINER_ID --timestamps > /cases/case-2024-001/docker/container_logs.txt 2>&1 # Capture running processes (if container is still running) docker top $CONTAINER_ID > /cases/case-2024-001/docker/container_processes.txt # Capture network connections docker exec $CONTAINER_ID netstat -tlnp 2>/dev/null > /cases/case-2024-001/docker/container_network.txt # Copy specific files from the container docker cp $CONTAINER_ID:/var/log/ /cases/case-2024-001/docker/container_var_log/ docker cp $CONTAINER_ID:/tmp/ /cases/case-2024-001/docker/container_tmp/ docker cp $CONTAINER_ID:/etc/passwd /cases/case-2024-001/docker/container_passwd # Hash all exported evidence sha256sum /cases/case-2024-001/docker/*.tar > /cases/case-2024-001/docker/evidence_hashes.txt
bash# Install dive for image layer analysis wget https://github.com/wagoodman/dive/releases/latest/download/dive_linux_amd64.deb sudo dpkg -i dive_linux_amd64.deb # Analyze image layers interactively dive forensic-evidence:case-2024-001 # Non-interactive layer analysis dive forensic-evidence:case-2024-001 --ci --json /cases/case-2024-001/docker/dive_analysis.json # Extract and examine individual layers mkdir -p /cases/case-2024-001/docker/layers/ tar -xf /cases/case-2024-001/docker/container_image.tar -C /cases/case-2024-001/docker/layers/ # List the image manifest and layer order cat /cases/case-2024-001/docker/layers/manifest.json | python3 -m json.tool # Examine each layer for changes for layer in /cases/case-2024-001/docker/layers/*/layer.tar; do echo "=== Layer: $(dirname $layer | xargs basename) ===" tar -tf "$layer" | head -20 echo "..." done # Use container-diff to compare with original base image # Install container-diff curl -LO https://storage.googleapis.com/container-diff/latest/container-diff-linux-amd64 chmod +x container-diff-linux-amd64 # Compare committed image with original ./container-diff-linux-amd64 diff daemon://nginx:latest daemon://forensic-evidence:case-2024-001 \ --type=file --type=apt --type=history --json \ > /cases/case-2024-001/docker/container_diff.json
bash# Docker data directory (default: /var/lib/docker/) DOCKER_ROOT="/mnt/evidence/var/lib/docker" # Examine overlay2 filesystem layers ls -la $DOCKER_ROOT/overlay2/ # Find the container's merged filesystem CONTAINER_HASH=$(docker inspect $CONTAINER_ID --format '{{.GraphDriver.Data.MergedDir}}' 2>/dev/null) # Or manually from forensic image: # Look in /var/lib/docker/containers/<container_id>/config.v2.json # Analyze container configuration files cat $DOCKER_ROOT/containers/$CONTAINER_ID/config.v2.json | python3 -m json.tool \ > /cases/case-2024-001/docker/container_config.json # Check Docker daemon configuration cat /mnt/evidence/etc/docker/daemon.json 2>/dev/null > /cases/case-2024-001/docker/daemon_config.json # Examine Docker events log cat $DOCKER_ROOT/containers/$CONTAINER_ID/*.log > /cases/case-2024-001/docker/container_json_logs.txt # Check for volume mounts (potential host filesystem access) python3 << 'PYEOF' import json with open('/cases/case-2024-001/docker/container_inspect.json') as f: data = json.load(f) inspect = data[0] if isinstance(data, list) else data print("=== CONTAINER SECURITY ANALYSIS ===\n") # Check mounts print("Volume Mounts:") for mount in inspect.get('Mounts', []): rw = "READ-WRITE" if mount.get('RW') else "READ-ONLY" print(f" {mount.get('Source', 'N/A')} -> {mount.get('Destination', 'N/A')} ({rw})") if mount.get('Source') in ('/', '/etc', '/var', '/root') and mount.get('RW'): print(f" WARNING: Sensitive host path mounted read-write!") # Check privileged mode host_config = inspect.get('HostConfig', {}) if host_config.get('Privileged'): print("\nWARNING: Container was running in PRIVILEGED mode!") # Check capabilities cap_add = host_config.get('CapAdd', []) if cap_add: print(f"\nAdded Capabilities: {cap_add}") dangerous_caps = ['SYS_ADMIN', 'SYS_PTRACE', 'NET_ADMIN', 'SYS_MODULE'] for cap in cap_add: if cap in dangerous_caps: print(f" WARNING: Dangerous capability: {cap}") # Check PID namespace if host_config.get('PidMode') == 'host': print("\nWARNING: Container shares host PID namespace!") # Check network mode if host_config.get('NetworkMode') == 'host': print("\nWARNING: Container shares host network namespace!") # Check user user = inspect.get('Config', {}).get('User', 'root (default)') print(f"\nRunning as user: {user}") # Check environment variables for secrets env_vars = inspect.get('Config', {}).get('Env', []) print(f"\nEnvironment Variables: {len(env_vars)}") for env in env_vars: key = env.split('=')[0] if any(s in key.upper() for s in ['PASSWORD', 'SECRET', 'KEY', 'TOKEN', 'CREDENTIAL']): print(f" SENSITIVE: {key}=***REDACTED***") PYEOF
bash# Compare container filesystem to original image docker diff $CONTAINER_ID > /cases/case-2024-001/docker/filesystem_changes.txt # A = Added, C = Changed, D = Deleted # Analyze changes python3 << 'PYEOF' added = [] changed = [] deleted = [] with open('/cases/case-2024-001/docker/filesystem_changes.txt') as f: for line in f: line = line.strip() if line.startswith('A '): added.append(line[2:]) elif line.startswith('C '): changed.append(line[2:]) elif line.startswith('D '): deleted.append(line[2:]) print(f"Files Added: {len(added)}") print(f"Files Changed: {len(changed)}") print(f"Files Deleted: {len(deleted)}") # Flag suspicious additions suspicious = [f for f in added if any(s in f for s in ['/tmp/', '/dev/shm/', '/root/', '.sh', '.py', '.elf', 'reverse', 'shell', 'backdoor'])] if suspicious: print(f"\nSuspicious Added Files:") for f in suspicious: print(f" {f}") # Flag suspicious changes sus_changed = [f for f in changed if any(s in f for s in ['/etc/passwd', '/etc/shadow', '/etc/crontab', '/etc/ssh', '.bashrc'])] if sus_changed: print(f"\nSuspicious Changed Files:") for f in sus_changed: print(f" {f}") PYEOF # Extract and examine the container export mkdir -p /cases/case-2024-001/docker/container_fs/ tar -xf /cases/case-2024-001/docker/container_export.tar -C /cases/case-2024-001/docker/container_fs/ # Scan for webshells and malicious files find /cases/case-2024-001/docker/container_fs/tmp/ -type f -exec file {} \; find /cases/case-2024-001/docker/container_fs/ -name "*.php" -newer /cases/case-2024-001/docker/container_fs/etc/hostname
bash# Scan the image for known vulnerabilities trivy image forensic-evidence:case-2024-001 \ --format json \ --output /cases/case-2024-001/docker/vulnerability_scan.json # Scan the exported filesystem trivy fs /cases/case-2024-001/docker/container_fs/ \ --format table \ --output /cases/case-2024-001/docker/fs_vulnerabilities.txt # Check for secrets in the image trivy image forensic-evidence:case-2024-001 \ --scanners secret \ --format json \ --output /cases/case-2024-001/docker/secrets_scan.json
| Concept | Description | |---------|-------------| | Image layers | Read-only filesystem layers stacked to form the container image | | overlay2 | Default Docker storage driver using union filesystem for layers | | Container diff | Comparison of runtime filesystem changes against the original image | | Privileged mode | Container with full host capabilities (bypasses most isolation) | | Docker socket | Unix socket (/var/run/docker.sock) controlling the Docker daemon | | Container escape | Technique for breaking out of container isolation to the host | | Volume mounts | Host filesystem paths made accessible inside the container | | Image history | Record of Dockerfile instructions used to build each layer |
| Tool | Purpose | |------|---------| | docker inspect | Detailed container configuration and state information | | docker diff | Show filesystem changes made in a running/stopped container | | dive | Interactive Docker image layer analysis tool | | container-diff | Google tool for comparing container image contents | | Trivy | Vulnerability scanner for container images and filesystems | | docker-explorer | Forensic tool for offline Docker artifact analysis | | Sysdig | Container runtime security monitoring and forensics | | Falco | Runtime threat detection for containers and Kubernetes |
Scenario 1: Web Application Container Compromise Export the container filesystem, identify webshells in web root, analyze access logs for exploitation attempts, check for added files and modified configurations, examine network connections for C2 communication, review container capabilities for escalation paths.
Scenario 2: Supply Chain Attack via Malicious Image Analyze image layers with dive to identify which layer added malicious content, compare with the official base image using container-diff, check image history for suspicious RUN commands, scan for embedded backdoors and cryptocurrency miners, trace the image pull from registry logs.
Scenario 3: Container Escape Investigation Check if container ran privileged or with dangerous capabilities, examine host filesystem mount points for unauthorized access, review Docker socket mount enabling Docker-in-Docker abuse, analyze host system logs for container escape indicators, check for kernel exploit artifacts.
Scenario 4: Cryptojacking in Container Environment Identify high-CPU containers, export and analyze the container image for mining binaries, check for unauthorized images in the registry, review container creation events for rogue deployments, examine network connections for mining pool communications.
Docker Container Forensics Summary:
Container: abc123def456 (nginx-app)
Image: company/web-app:v2.1
Status: Running (started 2024-01-10 09:00 UTC)
Host: docker-host-01.corp.local
Security Configuration:
Privileged: No
Capabilities Added: NET_ADMIN (WARNING)
Volume Mounts: /var/log -> /host-logs (RW)
Network Mode: bridge
User: root (WARNING)
Filesystem Changes:
Added: 23 files (5 suspicious)
Changed: 12 files (2 suspicious)
Deleted: 0 files
Suspicious Findings:
/tmp/reverse.sh - Reverse shell script (Added)
/var/www/html/.hidden/shell.php - PHP webshell (Added)
/etc/crontab - Modified (persistence cron entry added)
/root/.ssh/authorized_keys - Modified (unauthorized key added)
Vulnerability Scan:
Critical: 3 (CVE-2024-xxxx in base image)
High: 12
Medium: 34
Evidence: /cases/case-2024-001/docker/| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-21 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-08 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-22 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-11 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-12 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-01 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-10 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-03 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-16 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-07 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-02 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-06 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-19 | pass→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-17 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-04 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-05 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-09 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-13 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-14 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-15 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-18 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-20 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 21 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +9 percentage points is the difference between those two pass rates over the 21 comparable cases.
The per-case answers from this run were removed by the retention sweep, so the case table below shows the verdicts without the text either arm produced. The counts above were recorded at the time and are unaffected. Answers are now kept for 180 days.
Other measured skills in the registry, with their headline benchmark lift.