Use the heyvm CLI to log in, upgrade, create and manage local or cloud sandboxes, run commands inside VMs, deploy or update apps, expose/connect/share ports and shells, manage cloud databases, build VM images, and diagnose backend readiness.
Use and troubleshoot Heyo git submit CI/CD: install or upgrade the git-submit client, configure endpoints, submit branches, handle submodules, inspect CI runs, and clean up retained VM-backed jobs.
Use this skill any time you are running inside a heyvm sandbox (i.e. a `printer run` / `printer exec` turn dispatched through the heyvm sandbox driver). Explains what is mounted, what is ephemeral, and how to inspect the sandbox from inside it. Triggers when the agent needs to know "where am I", "is this state persistent", "what host paths are visible", or wants to check the sandbox's resources.
Use this skill whenever you need to modify source files in a repo where the codegraph tools are available. Apply changes by sending a unified diff to codegraph_patch instead of using edit/write to rewrite file contents. Triggers when the user asks to "edit a file", "modify code", "apply a patch", "make a change to <symbol>", or any time you would otherwise reach for edit/write on source code.
Use this skill to navigate and search a codebase efficiently with the codegraph tools instead of repeated greps and full-file reads. Triggers when the user asks to "find a symbol", "look up a definition", "list functions in a file", "show an outline", "find references", "search the codebase for X", or any time you need to locate code by name/signature/kind across a repo. Tree-sitter-backed; supports Rust, Python, JavaScript, and TypeScript.
Use this skill any time you are running as an ACP (Agent Client Protocol) server inside printer — that means poolside, opencode, claude-code-acp, or any other vendor with `kind = "acp"` in its plugin manifest. Explains the wire contract printer expects (one turn = one `session/prompt`, persistent session, cwd-is-live-filesystem) plus the project-conventions discipline that applies regardless of which model is driving. Skip vendor-specific quirks — those live in the per-vendor skill that ships alongside this one.
This skill should be used when an agent needs to drive or verify desktop/UI behavior on a Wayland session — during either implementation (exercising a UI/web change you just made: open a URL, click through flows, capture before/after screenshots) or review (confirming the change actually landed). Lists monitors and toplevel windows, captures screenshots, opens URLs in the default browser, and synthesizes keyboard/mouse/scroll input via uinput. Reach for this skill any time the user asks to "use computer", "screenshot the desktop", "list windows on Wayland", "click at coordinates", "type into the focused app", "send a key chord", "open a URL in the browser", or to confirm by visual evidence that a UI change actually works. Native Wayland desktop automation tool — use this when `$XDG_SESSION_TYPE` is `wayland`.
Use this skill when you are running as the opencode ACP agent inside printer (i.e. printer launched the `opencode` CLI in `acp` server mode via `--agent acp:opencode-acp`). Covers opencode-specific behavior — provider/auth via `opencode auth`, multi-provider support, configuration via `opencode.json` and `~/.config/opencode/` — that the generic `acp-runtime` skill doesn't cover. The runtime contract (one turn = one `session/prompt`, etc.) lives in `acp-runtime`; install that skill alongside this one.