Agent MCP Cheatsheet
Rogue Arena ships 190+ MCP tools driven by agent skills. The ones you can invoke directly are listed below. /architect-brainstorm runs design, YML checks, and canvas implement from sequential markdown, then you run /architect-final-validate.
Type the command to start a skill. Rogue skills don’t auto-trigger from conversation, so describing what you want won’t launch one. That’s deliberate, so a stray sentence can never kick off a build against a live canvas.
Active Deployment
Section titled “Active Deployment”Operate on a running deployment from your terminal.
| Command | What it does |
|---|---|
/rogue-active-deployment | List VMs, execute commands, read/grep files, list processes, manage snapshots, and transfer files. |
/rogue-maldev | Red-team malware development and detection testing loop: snapshot → exec → SIEM query → revert → iterate. Bring your own tool, a technique list, or research TTPs. |
Lab course tools (no command, just ask): read the lab’s course and hints, and submit a key you found; each submission uses one of your unlock attempts. See The MCP Toolkit.
Rogue Architect
Section titled “Rogue Architect”Design scenarios, audit canvases, debug deployments, and write curriculum.
| Command | What it does |
|---|---|
/architect-brainstorm | Design + validate + implement from one command (sequential markdown, no skill hops). Then run /architect-final-validate. |
/architect-final-validate | Final pre-deploy audit: plugin coupling, run-order, realism grade. Read-only. |
/architect-debug-deploy | Investigates deployment failures (Ansible errors, PowerShell issues, plugin misconfigurations). Requires a deployment ID. |
/architect-freeform | Manual canvas work with MCP tools. No guided pipeline; for ad-hoc adjustments to an existing canvas. |
/rogue-curriculum-builder | Build and edit curriculum: chapters, sections, CTF nodes, unlock keys, bulk inserts. Handles LINEAR and CTF layouts. |
Build management (no command, just ask): list canvases, read build issues, preview a redeploy plan, or remove a build. See Agents and Builds.
Eval authoring (no command, just ask): set up a Model Testing eval, write its hops, and dry-run them on the live build. See Build your own eval.
Plugin Dev
Section titled “Plugin Dev”Build and iterate on the Ansible plugins that machines run during builds.
| Command | What it does |
|---|---|
/rogue-plugin-brainstorm | Brainstorm a new Ansible plugin project: research install approaches, scaffold local files. |
/rogue-plugin-develop | Iterate on an existing plugin: write YAML, update vault files, maintain download scripts. |