Guide · Agent skill
How to use the Coercing Authentication with Coercer agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
This page explains the Coercing Authentication with Coercer skill, an agent skill that guides an AI agent through forcing a Windows machine account to authenticate to a listener the tester controls, using Coercer and PetitPotam. It is written for security professionals on authorized engagements and for blue teams checking their detections and mitigations. By the end you will know what the skill does, how to install it, how to phrase requests that trigger it, and what it needs before you use it.
Coercing Authentication with Coercermukul975/Anthropic-Cybersecurity-Skills · Apache-2.0 Download the skill Quick start
- 1Confirm authorization. Check that you have written permission that covers coercion and NTLM relay against the named hosts. Do not run the skill otherwise.
- 2Download and unzip the skill. Download the ZIP from the directory page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder.
- 3Or upload it in the Claude apps. Open Settings, go to Capabilities, then Skills, and upload the skill.
- 4Install the prerequisites. On a Linux attack host with Python 3.8+, install Coercer, get the PetitPotam source and install Impacket and Certipy, following the official project pages.
- 5Make a natural request. Ask for a scan of an authorized target using Coercer and state the scope and goal. The agent will load the skill from its description.
- 6Review the output. Read which RPC methods are reachable and decide with the client whether further testing is in scope.
What the skill makes the agent do
Some Windows RPC interfaces accept a UNC path and cause the receiving server to authenticate to it with its machine account. This skill teaches an agent to use that behaviour in a penetration test: it runs Coercer and PetitPotam scans and coercion against a target, usually a Domain Controller, and then covers how a relay chain can continue into AD CS Web Enrollment (ESC8), LDAP (RBCD) or SMB.
The skill is built around three Coercer modes. Scan probes which RPC methods are reachable, coerce triggers authentication, and fuzz varies UNC paths for research. The skill also includes a legal notice stating that it is for authorized testing only.
- Scan: list which coercion methods a target exposes
- Coerce: trigger machine account authentication to a listener
- Fuzz: research path variations (research use)
- Relay follow-up: ESC8, LDAP (RBCD) or SMB, as described in the skill
Install the skill (ZIP download or Claude upload)
Download the skill ZIP from the directory page. The skill folder contains SKILL.md, scripts/agent.py and two reference files, api-reference.md and standards.md.
For Claude Code, unzip the folder into ~/.claude/skills/ for a personal install, or into a project's .claude/skills/ folder to share it with one repository. In the Claude apps, open Settings, go to Capabilities, then Skills, and upload the skill there.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into .claude/skills/ inside the project
- Claude apps: Settings → Capabilities → Skills, then upload
Install the tools the skill needs
The skill does not ship the attack tools. Its prerequisites list a Linux attack host with Python 3.8 or newer, Coercer installed with pip or pipx, PetitPotam source from GitHub, Impacket, and Certipy for relay and certificate authentication. The skill's own prerequisites section gives the install lines for these tools; check them against the official project pages before running anything.
- Python 3.8+ on a Linux attack host
- Coercer (pip or pipx install)
- PetitPotam source from GitHub
- Impacket (ntlmrelayx, secretsdump)
- Certipy for relay and certificate authentication
How to trigger the skill with natural requests
Agents load a skill when a request matches its description. This skill's description covers coercion with PetitPotam and Coercer, relay into AD CS, LDAP or SMB, and validating coercion detections and signing or EPA mitigations. Name the tool and the goal, and state that the work is authorized. A request that is vague or unauthorized will not match the skill's intended use.
- Scan an authorized lab Domain Controller for coercible RPC methods using Coercer
- Explain which coercion methods are reachable on our test server and how to mitigate them
- Help me validate that our coercion detections fire during an approved purple-team exercise
- Check whether signing and EPA mitigations block an NTLM relay in our authorized test
What the skill needs before you start
Written authorization covering coercion and NTLM relay is the first requirement. The skill also expects valid domain credentials, which are often low-privileged, and network access to the target's RPC endpoints on port 135, the dynamic RPC range and port 445. A relay listener such as Certipy relay or Impacket ntlmrelayx must run on a reachable host.
The skill is not suited to readers without Active Directory and attack-host experience. The workflow assumes you can set up the relay, read the scan output and decide when to stop.
- Authorized scope that includes coercion and NTLM relay
- Valid domain credentials (often low-privileged)
- Network access to ports 135, dynamic RPC and 445
- A relay listener on a reachable host
- Linux attack host with Python 3.8+
Safety notes: scripts and network use
This skill runs scripts and uses the network. The bundled scripts/agent.py and the tools it drives send traffic to target hosts, so run it only against systems you own or have written permission to test. A coercion-relay chain against a Domain Controller can lead to full domain compromise, and running it without permission is illegal.
Use the scan mode first to see what is exposed without firing a full attack. Keep a record of targets, methods and time windows, and share results with the blue team so detections can be tuned.
- Confirm written authorization and scope before any run
- Start with scan mode to limit activity
- Keep logs of targets, methods and timing
- Coordinate with the blue team when validating detections
What people use it for
- Scanning an authorized test Domain Controller to list which coercion methods are reachable
- Checking whether patched servers still expose coercible RPC methods
- Generating coercion telemetry so a blue team can tune detections during an approved exercise
- Validating that signing and EPA mitigations block relay in an authorized test
Frequently asked questions
Is the Coercing Authentication with Coercer skill free to use?
The skill is published under the Apache-2.0 licence. The tools it relies on, Coercer, PetitPotam, Impacket and Certipy, are separate projects with their own licences, which you should check on their GitHub pages.
How do I install the coercer petitpotam agent skill?
Download the ZIP from the directory page and unzip it into ~/.claude/skills/ for a personal install or into a project's .claude/skills/ folder. In the Claude apps you can instead upload it under Settings, Capabilities, Skills.
What do I need to run this skill?
You need written authorization, valid domain credentials, network access to ports 135, the dynamic RPC range and 445, a relay listener, and a Linux attack host with Python 3.8 or newer. You also need Coercer, PetitPotam source, Impacket and Certipy installed.
What is the difference between Coercer scan, coerce and fuzz modes?
Scan probes which RPC methods on a target are reachable without launching a full attack. Coerce triggers the target machine account to authenticate to a listener. Fuzz varies UNC paths to research bypasses of partial patches.
How does PetitPotam compare with Coercer?
PetitPotam uses the MS-EFSR protocol only. Coercer is a broader tool that covers several protocols, including MS-EFSR, MS-RPRN, MS-DFSNM and MS-FSRVP, and offers scan, coerce and fuzz modes.
What should I do if the skill does not trigger?
Check that the skill folder sits directly in ~/.claude/skills/ or .claude/skills/, or that the upload succeeded in Settings, Capabilities, Skills. Then name Coercer or PetitPotam and state the authorized goal in your request.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you are a security professional running an authorized Active Directory engagement, or a blue team validating coercion detections and signing or EPA mitigations. Pick something else if you lack written permission for the target, or if you are new to Active Directory, NTLM relay and attack-host work, since the chain can lead to full domain compromise.
Download the skill