How to use the Linux Credential Pivot CTF agent skill

What is an agent skill, and what does this one do?

An agent skill is a folder of instructions and reference files that an AI agent loads when a request matches the skill's description. This one is a specialist add-on for Linux challenge nodes in CTF environments. It is not a general Linux guide.

Once the skill is active, the agent follows a structured method. It inventories credential artifacts, checks where each one is accepted, and records the chain from a recovered secret to the resulting access on another host.

• Maps SSH keys, agent sockets, kubeconfigs, cloud tokens, service-account secrets, environment variables and config files
• Notes sudoers rules, capabilities, setuid binaries, systemd unit context and namespace boundaries
• Shows where a key, token, socket or secret is accepted, such as SSH, an API, a Unix socket, a container runtime or a control-plane endpoint
• Separates authentication success from actual capability gain
• Reduces findings to one chain: recovered artifact, accepted replay path, pivot host or privilege change, resulting capability

What do you need before installing it?

The skill depends on another skill, ctf-sandbox-orchestrator. The downstream skill is meant to run only after that parent skill has set the sandbox assumptions, node ownership and evidence priorities and routed the task here. If the parent is not available, this skill is not the right starting point.

You also need a sandboxed CTF or lab environment in which the challenge nodes are under your control. The skill includes a bundled reference checklist, references/linux-credential-pivot.md, which the agent loads during work.

• Access to the ctf-sandbox-orchestrator skill
• A sandboxed CTF or lab environment with the challenge nodes under your control
• The bundled reference file references/linux-credential-pivot.md

How do you install the skill?

Download the ZIP file from this page and unzip it. Place the resulting folder in ~/.claude/skills/ to make it available to your user account, or in a project's .claude/skills/ folder to make it available only within that project.

If you use the Claude apps instead, you can upload the skill under Settings, then Capabilities, then Skills.

The package contains agents/openai.yaml, references/linux-credential-pivot.md and SKILL.md. Keep the folder structure intact so the agent can find the reference file.

• Option 1: unzip into ~/.claude/skills/ for personal use
• Option 2: unzip into a project's .claude/skills/ folder for one project
• Option 3: upload through Settings, then Capabilities, then Skills in the Claude apps

How do you trigger the skill with a natural request?

You do not need a special command. Describe the task in plain language, using the kinds of topics in the skill description. The agent should load the skill when your request matches it and the parent skill has already run. If the agent does not pick it up, first confirm that ctf-sandbox-orchestrator is active and that you have stated the sandbox context.

The skill's own instructions name the trigger topics: Linux auth artifacts, accepted token or key replay, socket or service-account trust edges, sudo or capability abuse, and lateral movement across Linux challenge nodes.

• Trace which SSH key or token is accepted by which host in our CTF lab
• Explain the trust edge between this Unix socket and the service account behind it
• Check whether this sudo rule or capability gives a path to another challenge node
• Reconstruct the lateral movement chain from the first foothold to the database host

What output should you expect?

The skill asks the agent to keep each artifact tied to its owner, scope and expected accepting service. It also asks for the exact pivot order, with protocol and target host or namespace, and one minimal replayable chain that proves the capability gain.

The skill's instructions say to reply in Simplified Chinese unless you explicitly request English. Ask for English in your request if you need it.

• Artifact path, owner, scope, accepting service and resulting principal
• Pivot order with protocol and target host or namespace
• One minimal replayable chain showing the capability gained

What are the safety limits?

This skill is built to find and replay live credentials and to pivot between hosts. Use it only on systems you own or have written permission to test, and only inside a sandboxed CTF or lab environment. The source lists real systems and networks you do not own, and production environments, as not suitable targets.

Replaying tokens or keys in production can lock accounts or cause outages. The skill's source states that it runs no scripts and uses no network access, but the agent's actions in your environment are still yours to control and supervise. If a chain moves into kernel exploit boundaries, the skill says to hand off to a kernel container escape skill instead of continuing here.

• Do not use it on systems you have no written permission to test
• Do not use it in production environments
• The skill runs no scripts and uses no network, according to its listing
• Hand off kernel-level chains to a separate kernel container escape skill

Tips for getting useful results

Start by stating your sandbox assumptions and which nodes you control, so the parent skill can establish them. The skill works best on one minimal path at a time. Separate local privilege escalation from lateral host pivot, because the reference checklist warns that mixing them in one vague chain is a common mistake.

Also check that the reference checklist and chain structure match what you are documenting, especially for write-ups.

• Give the node names, the account in use and the lab boundaries up front
• Ask for one chain at a time rather than a full map of the lab
• Keep a list of listed secrets that are proven accepted on a target, because listing them alone is not enough
• Ask for the evidence packaged as artifact, replay and effect, for a clean write-up