- Home
- Skills
- DevOps & Cloud
- Ransomware Incident Response Runbook Generator
Works with the AI Agents you already use
Ransomware Incident Response Runbook Generator
Generate stack-specific ransomware incident response runbooks with minute-by-minute tactical containment steps.
$5
Ransomware Incident Response Runbook Generator
Example session with this skill installed
- Tech Stack: AWS EC2, Windows Server 2022, Microsoft SQL Server, CrowdStrike Falcon, Veeam Backup - Company Size & Industry: 500 employees, Financial Services - Regulatory Jurisdiction: European Union (GDPR) and UK (FCA)
- Read your context and instructions
- Compiled the ransomware incident response
Ransomware Incident Response Runbook
Nov 1, 2026 · @Skill Ninja
Target Stack: AWS EC2, Windows Server 2022, Microsoft SQL Server, CrowdStrike Falcon, Veeam Backup |
Jurisdiction: EU (GDPR, DORA) and UK (FCA, UK GDPR) |
Profile: 500 employees, financial services
Before You Start
This runbook works for a 500-person financial firm running Windows Server 2022 and SQL Server on EC2, protected by CrowdStrike Falcon and backed up by Veeam. Execute Phase 1 in order. The regulatory clocks in Phase 2 start when you become aware of a likely breach, not when encryption is confirmed.
Assumptions (edit anything that is wrong; commands change if they are):
- SQL Server runs on EC2 instances, not Amazon RDS.
- Active Directory domain controllers are Windows Server 2022 (on EC2 or on-premises).
- Falcon sensors are on all servers, and your team holds Real Time Response (RTR) and containment permissions.
- Veeam Backup & Replication 12.x protects the servers, with a hardened (immutable) repository and a configuration backup. Confirm your exact version before using the Veeam steps.
- All commands need approval from the Incident Commander before running. Run them from a clean admin workstation, not from a suspect server.
Roles
| Role | Responsibility | Primary | Backup |
|---|---|---|---|
| Incident Commander (IC) | Owns decisions, the timeline and the incident log | [name / mobile] | [name / mobile] |
| Technical Lead | Runs containment, forensics and restoration | [name / mobile] | [name / mobile] |
| DPO / Legal Counsel | Breach assessment, regulator and data-subject notices | [name / mobile] | [name / mobile] |
| Compliance Officer (FCA SMF) | Regulator liaison, FCA / PRA notices | [name / mobile] | [name / mobile] |
| Communications Lead | Staff, client and media statements | [name / mobile] | [name / mobile] |
| Cyber Insurer / IR Retainer | Breach coach, forensics, negotiation advice | [24/7 hotline] | [policy no.] |
Out-of-band communication: Assume email, Teams and the domain are compromised. Move the response team to a pre-agreed channel (for example Signal or a separate tenant) at T+0, and keep one written incident log with timestamps in UTC.
Phase 1: Tactical Containment & Isolation (T+0 to T+60 Minutes)
Goal: stop the spread in 15 minutes, preserve evidence by minute 30, protect backups by minute 45, and decide the next move by minute 60. T+0 is the moment anyone credibly reports ransomware activity (Falcon ransomware detection, ransom note, mass file renames, or encrypted shares).
WARNING: Do not power off, reboot, snapshot-and-terminate or reimage any affected server yet. Encryption keys and attacker tooling may exist only in memory, and AWS cannot capture RAM from a stopped instance. Do not log in to affected servers with domain admin credentials; they can be harvested.
T+0 to T+15: Declare, Contain, Cut Identities
Declare the incident. The IC starts the incident log (UTC timestamps), moves the team to the out-of-band channel and notes the exact time of awareness. This time starts the GDPR and DORA clocks in Phase 2.
2.
Network-contain every affected host with CrowdStrike Falcon. In the console: Host setup and management, then Host management, select the hosts, then Network contain. A contained host is cut off from the network but still talks to the Falcon cloud, so Real Time Response keeps working. For bulk containment through the API with FalconPy:
from falconpy import Hosts
falcon = Hosts(client_id="<id>", client_secret="<secret>")
falcon.perform_action(action_name="contain", ids=["<host_aid_1>", "<host_aid_2>"])
Isolate at the AWS layer for any instance that has no sensor, an unresponsive sensor, or that is still spreading east-west. Swap the instance to a quarantine security group that has no inbound rules and no outbound rules (a new security group allows all outbound by default, so remove that rule):
aws ec2 modify-instance-attribute --instance-id i-0abc123example --groups sg-0quarantine123
WARNING: Security groups track connections. Already-established sessions can continue after the swap. If you need a hard cut, also add a deny rule in the subnet network ACL. A quarantine group with no outbound path to the Falcon cloud will also break Real Time Response, so prefer Falcon containment where the sensor is healthy.
Disable compromised identities. Disable the accounts seen in Falcon detections, and any privileged accounts that logged on to affected hosts, from a clean admin workstation:
Disable-ADAccount -Identity "<samAccountName>"
Revoke cloud credentials. Deactivate suspected IAM access keys and cut off active sessions for roles on affected instances:
aws iam update-access-key --user-name <user> --access-key-id <AKIA...> --status Inactive
For an IAM role, use Revoke active sessions in the IAM console (it attaches a deny policy for sessions issued before now).
T+15 to T+30: Preserve Evidence and Scope
- Snapshot every affected EBS volume (crash-consistent copy of disk state for forensics):
aws ec2 create-snapshots --instance-specification InstanceId=i-0abc123example --description "IR-<case-id> forensic" --copy-tags-from-source volume
- Capture volatile data through Falcon RTR before anything restarts. Typical read-only commands:
ps,netstat,ipconfig,eventlog(export Security, System and PowerShell logs) andget <path>to retrieve ransom notes and suspicious binaries. Have your IR retainer capture a full memory image; AWS has no native RAM capture. Confirm command names against your Falcon version.
Preserve logs off-host. Copy CloudTrail, VPC Flow Logs, GuardDuty findings, Windows event logs and SQL Server error logs to a write-once S3 bucket (Object Lock) in a separate forensic AWS account.
9. Hunt for cloud-side tampering. Look for attackers deleting backups or turning off logging:
aws cloudtrail lookup-events --lookup-attributes AttributeKey=EventName,AttributeValue=StopLogging
aws cloudtrail lookup-events --lookup-attributes AttributeKey=EventName,AttributeValue=DeleteSnapshot
aws cloudtrail lookup-events --lookup-attributes AttributeKey=EventName,AttributeValue=CreateAccessKey
Scope with Falcon. Search detections and host timelines for the first malicious process, the initial access host, and lateral movement (RDP, PsExec, WMI). Record patient zero and the earliest timestamp.
T+30 to T+45: Protect Backups and SQL Server
Freeze Veeam. Disable all backup jobs so encrypted data is not backed up over good restore points, and keep the Veeam server off the production network (Falcon-contain it if there is any doubt). On the Veeam server:
Get-VBRJob | Disable-VBRJob
Confirm restore points exist. In the Veeam console, check the hardened repository and the most recent clean restore points. Do not run retention or delete anything. Export a fresh configuration backup to offline media.
13. Check SQL Server for persistence, from a clean host with a break-glass login:
SELECT name, is_disabled FROM sys.server_principals WHERE IS_SRVROLEMEMBER('sysadmin', name) = 1;
SELECT name, value_in_use FROM sys.configurations WHERE name IN ('xp_cmdshell','Ole Automation Procedures');
SELECT TOP 20 database_name, backup_start_date, type FROM msdb.dbo.backupset ORDER BY backup_start_date DESC;
Disable unknown sysadmin logins with ALTER LOGIN [name] DISABLE;. If xp_cmdshell is on, turn it off (EXEC sp_configure 'show advanced options', 1; RECONFIGURE; EXEC sp_configure 'xp_cmdshell', 0; RECONFIGURE;).
WARNING: Do not detach databases, delete .mdf or .ldf files, or run DBCC CHECKDB with REPAIR_ALLOW_DATA_LOSS on affected instances. They destroy evidence and recoverable data.
T+45 to T+60: Decide and Escalate
Assess data theft. Check Falcon and VPC Flow Logs for large outbound transfers and archive tools. Treat the incident as a personal data breach until proven otherwise (double extortion is the norm).
15.
Decide whether to cut the environment off. If spread continues, sever internet and inter-VPC routes. This takes production offline, so only the IC may approve it:
aws ec2 replace-route --route-table-id rtb-0abc123example --destination-cidr-block 0.0.0.0/0 --gateway-id <blackhole-or-quarantine-target>
Activate the retainer and insurer. Notify the cyber insurer first if the policy requires consent before engaging vendors. Brief the executive team and the board chair.
17.
Hand over to Phase 2. The DPO and Compliance Officer start the notification clocks using the awareness time from step 1.
Ransom policy: Do not contact the threat actor or pay. Any contact goes through legal counsel and the IR retainer, after sanctions checks (UK OFSI and EU sanctions lists). A decryptor is never a substitute for rebuilding.
Phase 2: Assessment & Legal Notification Protocols
Four regulatory clocks can run at once, and the earliest is DORA's 4-hour initial notification. The DPO and Compliance Officer own them from T+45 minutes. Counsel should direct the forensic work so that findings are privileged. Notify with what you know and update later; a late notice costs more than an incomplete one.
Notification Timeline
| Obligation | Notify | Deadline | Clock starts | Owner |
|---|---|---|---|---|
| DORA Art. 19 initial notification (EU entities in scope) | National competent authority, via its incident platform | As early as possible within 4 hours of classifying the incident as major, and no later than 24 hours after becoming aware | Classification as major; 24-hour backstop from awareness | Compliance Officer |
| DORA intermediate report | Same authority | Within 72 hours of the initial notification, or when regular activity has recovered; update on significant status changes | Initial notification | Compliance Officer |
| DORA final report | Same authority | Within one month of the latest intermediate report | Latest intermediate report | Compliance Officer |
| EU GDPR Art. 33 | Lead supervisory authority (main establishment), else each affected member-state authority | Without undue delay and within 72 hours of becoming aware, in phases if needed (Art. 33(4)) | Awareness of a personal data breach | DPO |
| UK GDPR | Information Commissioner's Office (ICO) | Within 72 hours of becoming aware, where feasible | Awareness | DPO |
| GDPR Art. 34 (EU and UK) | Affected individuals | Without undue delay, if the breach is likely to result in high risk to them | Risk assessment | DPO, Comms |
| FCA Principle 11 and SUP 15.3 | Your named FCA supervisor, or the SUP 15 notification form; also the PRA if dual-regulated | Promptly. The FCA guidance reviewed sets no fixed hour count, so treat same day as the standard | Awareness of a material incident | Compliance Officer (SMF) |
| Clients and counterparties | Affected clients, per contract and DORA Art. 19(3) where their financial interests are affected | Without undue delay; follow contract notice terms | Impact confirmed | Comms, Legal |
| Cyber insurer | Insurer hotline | Per policy (often immediate, before engaging vendors) | T+0 | Incident Commander |
| Law enforcement and NCSC | UK police and NCSC; EU national police and CSIRT | As soon as practical (supports claims and intelligence) | Scoping | Legal |
Status check, 1 Nov 2026: New FCA and PRA operational incident reporting rules (PS26/2) take effect on
18 March 2027. They are not yet in force, so Principle 11 and SUP 15.3 apply today. Build the new single-form process now. DORA applies only to the EU financial entities in your group; confirm which legal entity each affected system belongs to. This runbook is operational guidance, not legal advice; have counsel confirm deadlines for your entities.
Assessment Steps (T+1 hour to T+24 hours)
Classify for DORA. An incident is major if critical or important functions are adversely affected and either there is successful malicious unauthorised access that may cause data loss, or two or more materiality thresholds are met (clients, transactions, duration, geography, data, economic impact). Record the time of classification, because the 4-hour clock runs from it.
2.
Assess the personal data breach. Establish which databases and file shares were reached, the data categories (customer financial data, identity documents, special category data, employee data), the number of individuals, their countries, and whether data was encrypted only or also exfiltrated. Review SQL Server database names against your data inventory.
3.
Record the risk decision. Write down whether notification to authorities and individuals is required and why. Keep this in the breach register even if you decide not to notify (GDPR Art. 33(5)).
4.
Open the FCA and PRA line. The Compliance Officer calls the named supervisor and follows with the written notice.
5. Issue a legal hold. Stop deletion of email, chat, logs and tickets related to the incident.
Forensic Preservation Rules
- Chain of custody: Log who collected each artifact, when, where it is stored, and its SHA-256 hash.
Keep originals untouched: Work from copies of EBS snapshots; keep the first snapshots read-only in the forensic account.
- Retain ransom notes, binaries and the attacker's communications, and note any Tor or negotiation URLs without visiting them.
- Do not rebuild or wipe any host until the IR retainer confirms its image and memory capture is complete.
- Time sync: Record each host's clock offset; align all timelines to UTC.
- Retention: Keep the evidence set until counsel and regulators confirm it can be released.
Communication Templates
1. Regulator initial notice (adapt to the authority's form)
Subject: Initial notification of major ICT incident / personal data breach: [Firm legal entity]
1. Entity and contact: [legal entity, LEI/FRN, 24/7 contact]
2. Time of detection: [UTC]; time of classification: [UTC]
3. Nature: Ransomware affecting [systems/services]; encryption confirmed on [n] servers.
4. Services and customers affected: [description, estimated number, countries]
5. Personal data involvement: [confirmed / suspected / under assessment]; categories: [list]
6. Containment taken: Affected hosts network-contained; credentials revoked; backups protected.
7. Impact on other entities or third parties: [description]
8. Next update: [time], or sooner on material change.
2. Staff holding statement (via out-of-band channel)
We are responding to a cyber security incident affecting some of our systems. Do not log in to company systems, open unexpected attachments or contact anyone claiming to be responsible. Use [alternative channel] for updates. Report anything unusual to [contact]. Next update at [time].
3. Client holding notice
We are investigating a cyber security incident that may affect [service]. Our priority is protecting your information and restoring service. We have notified our regulators and will update you by [time]. If we confirm your data was affected, we will contact you directly. Contact: [number/email]. Please be alert to phishing messages that reference this incident.
Phase 3: Backup Verification & Clean Restoration
Restore only from verified-clean restore points, into an isolated network, as new instances, with identity rebuilt first. Never overwrite a compromised host and never reconnect a restored server to production until it passes the gate for its tier. Expect the first critical services back in 24 to 72 hours; set realistic targets with the business using your impact tolerances.
Step 1: Build the Clean Room (start in parallel with Phase 2)
Create an isolated VPC or subnet with no internet gateway, no peering and no transit gateway attachment to production. Allow outbound access only to the Falcon cloud, the Veeam repository and a patch source.
2.
Prepare a clean admin workstation (freshly built, Falcon installed) and new admin accounts. Do not use any pre-incident administrator credential.
3.
Rotate cloud secrets before restoring: IAM keys and roles, KMS grants where access was abused, Falcon API client secrets and Veeam stored credentials.
Step 2: Choose the Restore Point
- Take the earliest confirmed compromise time from Phase 1, then subtract a safety margin set by the IR retainer (often weeks, because attackers dwell before encrypting).
- List restore points on the Veeam server and pick the newest one that predates that time:
$b = Get-VBRBackup -Name "<backup job name>"
Get-VBRRestorePoint -Backup $b | Sort-Object CreationTime -Descending | Select-Object Name, CreationTime, Type
- Confirm the restore point sits on the hardened (immutable) repository, was not modified after creation, and is covered by an intact configuration backup.
WARNING: If the Veeam server itself was compromised, do not restore from it as it stands. Rebuild a new Veeam server in the clean room and import the repository and a pre-incident configuration backup.
Step 3: Verify Backup Integrity and Cleanliness
| Check | How | Pass criterion |
|---|---|---|
| Malware-free image | Restore with Veeam Secure Restore (antivirus scan during restore; YARA scanning where your version supports it), or restore into the clean room and scan | No detections in Veeam or Falcon |
| Bootable and consistent | Restore to a new EC2 instance in the isolated subnet; if you use SureBackup for your image-level backups, run it for the critical servers | Boots; services start; no errors |
| Falcon coverage | Install the sensor on the restored host before it joins any network | Host visible in the console; prevention policy applied |
| Compromise hunting | Search the restored host for attacker tools, persistence (scheduled tasks, services, WMI subscriptions) and unknown local admins | None found; findings go to the IR retainer |
| SQL Server data | Verify native backups, then check the restored database | RESTORE VERIFYONLY and DBCC CHECKDB both clean |
WindowsSensor.exe /install /quiet /norestart CID=<your-CID-with-checksum>
RESTORE VERIFYONLY FROM DISK = N'D:\restore\<db>.bak' WITH CHECKSUM;
DBCC CHECKDB (N'<RestoredDB>') WITH NO_INFOMSGS, ALL_ERRORMSGS;
Step 4: Rebuild Identity First
Restore or rebuild domain controllers in the clean room from pre-compromise points, then check AD for rogue accounts, group changes, GPOs, scheduled tasks and trust changes.
2.
Reset the krbtgt password twice, waiting at least the maximum ticket lifetime (10 hours by default) between resets. Use Microsoft's published krbtgt reset script rather than a manual change.
3. Reset all privileged, service and SQL Server service-account passwords, rotate sa and gMSA secrets, then force a password reset and MFA re-registration for all users.
Step 5: Restore Services in Tier Order
| Order | Tier | Gate before connecting to production |
|---|---|---|
| 1 | Domain controllers, DNS, certificate services | Clean AD review complete; krbtgt reset twice; Falcon healthy |
| 2 | Veeam server, logging, monitoring | New credentials; repository immutability confirmed |
| 3 | SQL Server instances | Database checks clean; sysadmin list reviewed; xp_cmdshell off |
| 4 | Critical and important business applications (payments, trading, client portals) | Business owner sign-off; reconciliation plan agreed (Step 6) |
| 5 | Remaining servers, then user access in waves | 7 days of clean monitoring on tiers 1 to 4 |
Step 6: Reconcile Financial Data
- Calculate the gap between the restore point and the incident: transactions, instructions and records created after the backup.
- Rebuild that gap from upstream and downstream sources (payment gateways, custodians, counterparties, message logs, email). Have Finance and Compliance sign off on the reconciliation.
- Record the data loss window; you will need it for the DORA final report and any regulator or client questions.
Step 7: Business Continuity While Systems Are Down
Run manual or alternative procedures for each critical or important business service, using the pre-agreed impact tolerances.
- Hold fixed update times for the executive team, regulators and clients, even when there is nothing new to say.
Prioritise client-facing and market-facing obligations (settlement cut-offs, regulatory filings) and warn counterparties early if any will be missed.
Step 8: Stabilise and Learn
- Keep the heightened Falcon prevention policy and daily threat hunting on for at least 30 days.
- Submit the DORA final report within one month of your latest intermediate report, and complete ICO and supervisor follow-ups.
- Hold a lessons-learned review within two weeks, and update this runbook, the asset inventory and the backup design (immutability, separate credentials, tested restores).
Critical Success Factors
Most failed recoveries come from a few repeated mistakes: acting before preserving evidence, restoring too early, and letting the attacker's access survive.
Never reboot, power off or wipe before capture. Volatile evidence and decryption clues disappear, and containment through Falcon and security groups keeps servers alive but harmless.
Start the clocks at awareness. DORA's 4-hour initial notification and the 72-hour GDPR limit run in parallel; a documented partial notice beats a late complete one.
Protect backups before anything else is restored. Attackers target Veeam servers, repositories and cloud snapshots first. Keep an immutable copy and separate credentials.
Assume credentials are stolen. Rebuild identity first, reset krbtgt twice with the required wait, and rotate every service and cloud secret before reconnecting servers.
Restore into isolation, as new instances. Never overwrite a compromised host, and never join a restored server to production until it passes its tier gate.
- Choose the restore point from the compromise date, not the encryption date. Dwell time is usually long. Restoring a point that already contains the attacker re-infects you.
Verify every restore. Run malware scans, DBCC CHECKDB and a Falcon health check, and reconcile financial data before business owners sign off.
Keep one incident log and one voice. The IC approves every destructive command; Legal and Comms approve every external statement.
Treat the ransom as a legal question, not a technical one. Sanctions checks, counsel and the insurer come first; payment does not guarantee data return or deletion.
Test this runbook. Run a tabletop at least twice a year, including a Veeam restore to an isolated VPC, and refresh contacts and commands after every infrastructure change.
Sources
- FCA: Reporting operational incidents: current Principle 11 and SUP 15.3 position, PS26/2 effective 18 March 2027, ICO 72-hour reference.
- Austrian FMA: DORA ICT-related incidents: DORA major-incident classification and the 4-hour, 24-hour, 72-hour and one-month reporting deadlines.
GDPR Article 33 and 34 timings come from the regulation text. CLI commands and console paths follow vendor documentation as I know it; confirm Falcon, Veeam and AWS syntax against your installed versions in a test environment before an incident.
Connects securely to your tools. The creator never sees your data.
About this skill
Cybersecurity incidents move faster than manual documentation can keep up. This skill generates stack-specific, minute-by-minute ransomware response runbooks for security planning, compliance audits, or crisis simulations. It replaces generic templates with precise technical commands and regulatory-specific notification schedules.
What it does
- Tactical Containment provides chronological, minute-by-minute isolation steps for specific cloud and infrastructure providers.
- Forensic Preservation details methods to isolate affected assets without destroying volatile evidence.
- Regulatory Mapping identifies mandatory reporting timelines and notification protocols based on specific jurisdictions.
- Restoration Workflows outlines step-by-step procedures for verifying backup integrity and executing clean production restores.
How it works
- Provide the specific technology stack, including cloud providers, database engines, and virtualization layers.
- Define the company size, industry, and governing jurisdictions like GDPR, CCPA, or HIPAA.
- The skill generates a three-phase operational runbook covering containment, legal notification, and restoration.
Frameworks & tools
This skill supports infrastructure providers like AWS, Azure, and GCP, along with container orchestration, legacy on-premise virtualization, and major backup solutions.
Why this beats prompting it yourself
Standard prompts often produce generic advice that lacks technical depth. This skill enforces a zero-hallucination constraint on infrastructure commands and applies precise chronological intervals to the response timeline, ensuring the output is an actionable consulting deliverable rather than a theoretical guide.
Use cases
- Generate emergency response playbooks for SOC teams during blue-team exercises.
- Create technical disaster recovery documentation for SOC2 or ISO 27001 compliance.
- Build stack-specific isolation guides for rapid onboarding of incident responders.
- Prepare legal and regulatory notification templates for specific regional mandates.
Known limitations
This skill provides guidance for planning and simulations. It cannot execute commands on your infrastructure or directly contact law enforcement.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for DevOps & Cloud.
Cannot execute infrastructure commands directly. Requires accurate tech stack input for technical accuracy. Does not provide real-time threat intelligence.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 16 days ago