How to use the Competition Mailbox Abuse agent skill

What the skill makes the agent do

The skill is documentation and workflow guidance for an AI agent. It is not a program that changes mail systems by itself. Its job is to help the agent follow a fixed method: map the mail trust path, prove the mailbox effect, and reduce the result to the shortest chain of events that explains it.

In practice the agent is told to keep identity evidence, mailbox evidence and message-trace evidence tied to the same user, mailbox, token or message ID. It records forwarding targets, rule conditions, consent scopes and shared mailbox edges, then writes up the smallest sequence that can be replayed.

• Traces consent grants, app registrations, sign-in events and token claims
• Reviews inbox rules, forwarding settings, delegate access, shared mailbox permissions and transport rules
• Correlates message traces and mailbox audit events by message ID, sender or recipient pair and timestamp
• Summarises the chain as lure or grant, then token or delegate edge, then mailbox or transport change, then mail effect

Check the requirements before you install

This skill only works as a downstream add-on. The parent skill ctf-sandbox-orchestrator must already be installed and active, and it must have set up the sandbox assumptions before this skill is used. You also need a sandbox or lab environment that contains mailbox and log data to analyse.

The skill does not run scripts and does not use the network. Its files are agents/openai.yaml, references/mailbox-abuse.md and SKILL.md.

• Parent skill ctf-sandbox-orchestrator installed and active
• A sandbox or lab with mailbox and log data
• Use only in an authorised CTF or training environment

How to install the competition-mailbox-abuse skill

Download the ZIP file from this page and unzip it. You can install the skill in one of two places. For a personal setup, put the unzipped skill folder inside ~/.claude/skills/. For a single project, put it inside that project's .claude/skills/ folder.

In the Claude apps you can instead upload the skill under Settings, then Capabilities, then Skills. Install the parent skill ctf-sandbox-orchestrator as well, because this skill will not be used without it.

How to trigger the skill with natural requests

You do not need a special command. The skill is chosen when your request matches its description. The description says it applies when you ask to trace mailbox rules, OAuth consent grants, forwarding or delegate abuse, shared mailbox access, message-trace evidence, or how mail artifacts lead to persistence, exfiltration or privilege.

The skill is meant to be used after the parent skill has been activated and has routed the task here. Mention the parent skill in your first message to make this clear, for example by naming ctf-sandbox-orchestrator as the active workflow.

• Trace the inbox and forwarding rules for the user in this lab mailbox and list the forward targets.
• Correlate the consent grant logs with the sign-ins for this app and show which scopes were granted.
• Using the message trace export, follow message ID values from the shared mailbox and explain the resulting mail flow.
• Explain how this OAuth consent could lead to persistence in the lab tenant.

What a good result looks like

The skill asks the agent to keep its output compact. A useful answer names each artifact, such as the consent scope, rule predicate, forward target or transport action, and links it to a message ID or timestamp. It also states whether persistence sits in consent, mailbox rules, transport configuration or shared mailbox permissions.

The agent is told to avoid reporting a token or delegate edge without showing the mailbox effect it enables. It is also told not to attribute message flow to inbox rules or transport rules unless the evidence shows which one caused it.

• Consent scopes, token claims, rule definitions and forwarding targets listed with their message IDs
• A short replayable sequence that proves persistence, exfiltration or delegate access
• Clear separation between possessing a token and the mailbox effect it produces

Safety notes: scripts, network and authorisation

The skill runs no scripts and uses no network. It contains guidance only, so it will not scan your environment or send mail. Its instructions focus on reproducing abuse effects, which is why the intended use is a CTF lab or a training exercise.

Do not use this skill against real mailboxes or organisations without written authorisation. Keep all work inside the sandbox that the parent skill has defined.

• No executable scripts are included
• No network use is described by the skill
• Use only in authorised CTF labs, training exercises or defensive investigations

Tips for getting reliable results

Give the agent the logs and exports it needs before asking for an analysis. The skill depends on consent logs, sign-in records, message traces and mailbox audit events being present in the lab. Without them the agent cannot show the mailbox effect.

Keep one user, mailbox or message ID per question. Separate questions about inbox rules from questions about transport rules, because the reference warns that mixing them makes the cause hard to prove.

• Provide the lab's log and message trace data up front
• Ask about one user, mailbox or token at a time
• Ask for the smallest replayable chain rather than a full report
• Check that the parent skill is active if the agent drifts into host or Kerberos topics, which the skill says should go back to the broader identity skill