.agents

Use when the user wants to live-test an Android APK end-to-end on an emulator/device and get a works/doesn't-work + UI/UX report — driving the real app, not writing test code. Triggers: "test my Android app", "QA this APK", "run the app on the emulator and tell me what breaks", "click through every screen", "review the app's UX", "does my APK work", "test the built APK". Installs/launches the APK, enumerates screens from the accessibility tree, exercises each feature via adb (tap/type/swipe/launch), watches logcat for crashes/ANRs, captures screenshots + UI dumps, and emits a structured report. Primary path is direct local adb; optional richer path via claude-in-mobile. Sibling of web-app-testing and desktop-app-testing (shared report format). Does NOT fire for: building/coding an Android app (use claude-android-ninja / jetpack-compose-expert), iOS, or unit tests.

Use when the user wants to live-test a built Windows desktop application (.exe) end-to-end inside the agent-os Windows 11 VM and get a works/doesn't-work + UI/UX report — launching the real binary and driving it, not writing test code. Triggers: "test my Windows app", "QA this .exe", "run my desktop build on agent-os and tell me what breaks", "click through the app", "review the desktop app's UX", "does my exe work", "test the built binary in the Windows VM". Gets the .exe into the VM, launches it, enumerates controls from the UI Automation tree, drives every feature (click/type by element), watches for crashes/hangs/error dialogs, captures screenshots + control-tree dumps, and emits a structured report. Drives the agent-os VM via the Windows-MCP gateway. Sibling of web-app-testing and android-app-testing (shared report format). Does NOT fire for: building/coding a desktop app, the user's personal Windows on steamy (use nircmd), or general agent-os VM driving (use agent-os).

Use when the user wants to live-test a WEB app/site end-to-end in a real browser and get a works/doesn't-work + UI/UX report — not just write Playwright code. Triggers: "test my web app", "QA this site", "run E2E on the web app", "click through every feature and tell me what breaks", "review the UX of this web app", "does my web app work", "test the deployed site". Drives a real Chrome via Playwright over CDP (default http://127.0.0.1:9222), enumerates features from the DOM/ARIA tree, exercises each, captures screenshots + console/page/network errors, and emits a structured report. Sibling of android-app-testing and desktop-app-testing (shared report format). Does NOT fire for: writing a one-off Playwright script (use webwright), pure web design/build with no testing, or backend/API-only testing.

Read or interact with the broadcastr activity feed across concurrent Claude sessions. Use when the user says "what are other agents doing", "show recent activity", "mute plan-exec notifications", "what's on the bus", "broadcastr status", "tail broadcastr", "emit a manual broadcastr event", or asks why a notification appeared. Also use to disable broadcastr in the current session (BROADCASTR_DISABLED=1) or check the global feed. Does NOT fire on generic "show me activity" or "what's happening" unless broadcastr is explicitly named.

Use when the user asks to "work it", execute a superpowers executing-plans document in a worktree, create a PR as soon as the implemented plan is green, or run a complete review-and-fix loop over all touched files.

Use when the user asks to download, archive, search, inspect, or organize media with yt-dlp, MeTube, YouTube, playlists, channels, supported video sites, audio extraction, metadata, thumbnails, subtitles, Plex yt-dlp libraries, or NAS media downloads.

Save session documentation to a markdown file with full context — date, branch, HEAD, session ID, and git state pre-injected — then stage, commit, and push only the generated session artifact. Use when the user says "save session", "save to md", "document this session", "write up what we did", "save session notes", or asks to capture the current conversation as a session log. Pass `--html` or a `.html` path to render a rich Aurora-styled HTML artifact instead.

Install broadcastr's git-hook shims into the current repo. Use when the user says "install broadcastr hooks", "wire up broadcastr in this repo", "set up the broadcastr git hooks", or after first installing the broadcastr plugin in a new repo. Idempotent — safe to run repeatedly. Preserves any pre-existing hook as `<hook>.broadcastr-prev` and chains to it.

Create the save-to-md session doc before staging, git add all, commit with Claude co-authorship trailer, and push to current/new feature branch — including version bump and changelog update. Use when the user says "quick push", "push my changes", "commit and push", "ship this", "push to a new branch", or any request to wrap up local work and get it on the remote. Accepts optional `--no-bump` argument to skip the version bump.

Use when installing, setting up, repairing, or initializing the Plexus remote-device memory plugin; when the user asks where Plexus stores persistent REMOTE.md host profiles; or when seeded host profiles need to be created from bundled templates. Runs Plexus bootstrap to copy missing templates into plugin data without overwriting user-authored profiles.

Use when the user mentions operating on a named remote host that has a Plexus `remotes/<host>/REMOTE.md` profile, such as squirts, dookie, steamy, or another homelab device. Before taking action, load durable host memory and live context with `remote-context.py <host>`. This is the host-scoped equivalent of CLAUDE.md/AGENTS.md for remote machines.

Direct Dozzle API workflow for the real-time Docker container log viewer. Use when the user wants to inspect Dozzle health, version, containers, logs, authentication failures, session-cookie refresh, auth setup, or Dozzle MCP. Uses `DOZZLE_URL` and optional `DOZZLE_SESSION_COOKIE`; do not route through Lab MCP or `lab dozzle`.

Use when the user wants to use claude-in-mobile, its MCP server, or its native CLI to automate Android devices/emulators, iOS Simulators, Aurora OS devices, macOS desktop apps, Chrome/Chromium CDP sessions, screenshots/logs, UI inspection, accessibility or visual checks, multi-device flows, or app-store release operations.

Validate MCP-UI / MCP Apps implementations with MCPJam CLI and Inspector. Use when testing a server's UI-capable MCP tools, `_meta.ui.resourceUri`, `ui://` resources, `text/html;profile=mcp-app` resource contents, `structuredContent`, Inspector rendering, or Axon-style MCP Apps widgets.

Lab's Adguard integration — DNS-level ad blocking and network filtering. Use when the user wants to manage their Adguard instance, or invokes `lab adguard` / `mcp__lab__adguard`. Calls the MCP tool first, falls back to the CLI if MCP is unavailable.

Use to drive Claude's sandboxed Windows 11 VM, the `agent-os` VM, container `agent-os-win11` (`dockur/windows`) on host `dookie`, historically nicknamed "winbox", via the Windows-MCP server installed inside it. Triggers: "agent-os", "the agent-os VM", "winbox", "the windows sandbox", "the dookie windows", "drive the windows VM", "spin up agent-os", "run X on agent-os", "screenshot agent-os", "PowerShell on agent-os", or explicit noVNC at http://dookie:8006. For web-dev/browser verification prefer: CDP on agent-os, agent-browser, claude-in-chrome on agent-os, agent-os Windows-MCP, then claude-in-chrome on steamy. Prefer `mcp__windows-mcp__*` for desktop/OS work. Does NOT fire on the user's personal Windows on steamy-wsl; that target uses the `nircmd` skill. The sandbox is Claude's alone: install software, change settings, run shells freely without confirmation.

Lab's Apprise integration — universal notification dispatcher — 100+ services behind one URL scheme. Use when the user wants to manage their Apprise instance, or invokes `labby apprise` / `mcp__lab__apprise`. Calls the MCP tool first, falls back to the CLI if MCP is unavailable.

This skill should be used when the user asks to manage Docker containers, images, volumes, networks, stacks, environments, or mentions Arcane, Docker management, container operations, image updates, or GitOps workflows.

Use whenever building, modifying, or styling React/Next.js UI for any Aurora, Labby, or Lab/gateway-admin surface, or when the user mentions Aurora, Labby UI, the operator console, the aurora-design-system repo (`~/workspace/aurora-design-system`), `aurora.tootie.tv`, the `@aurora` shadcn registry, `registry/aurora/`, `--aurora-*` CSS tokens, `aurora-page-shell`, or asks for the dark operator/agent control-plane look. Make sure to invoke this skill any time React, Next.js, shadcn, Tailwind, or "dashboard"-style UI work happens inside `~/workspace/aurora-design-system` or in a project that consumes the Aurora registry, even if the user doesn't say "Aurora" by name. Aurora is dark-first, uses a navy palette with cyan primary / rose secondary / violet AI accents, muted status colors, a Manrope + Inter + JetBrains Mono type stack, and `var(--aurora-*)` semantic tokens — never raw hex.

Use whenever the user mentions Bitwarden — vault items, logins, passwords, secure notes, cards, identities, folders, collections, attachments, Sends, organizations (members, groups, policies, collections, events, subscription, device approvals), password/passphrase generation, vault sync/lock/unlock, BW_SESSION handling, the Bitwarden CLI (`bw`), or the Bitwarden MCP server. Prefer the MCP tools (`mcp__bitwarden__*`) over shelling out to `bw`; only fall back to the CLI for session lifecycle (unlock/lock/status) and operations the MCP server does not expose.

Manage code snippets in ByteStash snippet storage service. This skill should be used when the user asks to "save a snippet", "search snippets", "find code", "share snippet", "organize snippets", "list my snippets", "create snippet", "delete snippet", or mentions ByteStash, code storage, snippet management, or code archival.

Audit CLI dependencies referenced by active, installed, and disabled skills for Claude, Codex, and Copilot. Use when the user asks to check whether skill-required commands are installed, verify all skill CLIs, compare enabled versus disabled skill dependencies, or produce a report of missing, shadowed, or misresolved command-line tools used by local agent skills.

Use when the user wants to inspect or control a real Chrome session over SSH via Chrome DevTools Protocol. For web-dev/browser verification, prefer this ladder: CDP on agent-os, agent-browser, claude-in-chrome on agent-os, agent-os Windows-MCP, then claude-in-chrome on steamy. Triggers imply a real Chrome session: "grab my chrome tab", "show me my tabs", "screenshot my <site> tab", "what's open in my chrome", "eval this in my browser", "cookies from my chrome", "navigate my chrome to", "what's the page console showing", "check my chrome network requests". For the user's personal Chrome on steamy, require Chrome launched with `--remote-debugging-port=9222`; for generic automation, fall back instead of stopping.

Use when building Android apps with Kotlin, Jetpack Compose, MVVM, Hilt, Room 3 (KSP, SQLiteDriver, Flow/suspend DAOs), and multi-module architecture. Triggers on requests to create Android projects, modules, screens, ViewModels, repositories, or Android architecture questions. Not for iOS, Flutter, React Native, KMP-only shared code without an Android app module, or backend-only APIs with no Android client.

Use when the user wants to push text to or read text from their Windows clipboard over SSH. The killer use case is "copy this to my clipboard" — the agent can't paste images into chat over SSH, but it can shove URLs, commands, code snippets, multi-line text, Unicode, or anything else onto the user's Windows clipboard for Ctrl+V wherever they want it. Triggers include "copy this to my clipboard", "put X on my clipboard", "push this to clipboard", "set my clipboard to", "what's on my clipboard", "read my clipboard", "clipboard contents". Always targets `ssh steamy-wsl` (Jacob's primary Win11 desktop, where he works 99% of the time) — invoke this skill regardless of what host this Claude session is running on. Auto-routes between NirCmd (fast ASCII) and PowerShell `Set-Clipboard` via a UTF-8 temp file (lossless Unicode); degrades gracefully when NirCmd is absent.

Use whenever the user wants to add or scaffold a new reverse proxy config for SWAG — the LinuxServer.io SWAG proxy on `squirts` (`/mnt/appdata/swag/nginx/proxy-confs/`) that fronts every `*.tootie.tv` subdomain. Triggers: "create a swag config", "add a swag proxy for X", "add X to swag", "make a subdomain config", "expose X on tootie.tv", "add reverse proxy for X", "new tootie.tv subdomain", "proxy X through swag", "scaffold a swag entry", "wire up a SWAG mcp config". Always prefer the `swag-mcp` MCP server (registered as `swag` at `https://swag.tootie.tv/mcp`, exposes one action-routed tool: `swag` with actions list/create/view/edit/update/remove/logs/backups/health_check) over hand-writing files — it validates inputs, renders the canonical Jinja2 template, writes the file with the correct name, manages backups, and exposes its own health_check. Hand-write only when swag-mcp is unreachable. Does NOT trigger for nginx work outside this homelab.

Query MCP servers from the shell using fastmcp discover, list, and call. Use when you need to discover configured servers, list tools/resources/prompts, call tools, read resources, get prompts, or bridge MCP servers into shell-based workflows.

Lab's Freshrss integration — Self-hosted RSS feed aggregator. Use when the user wants to manage their Freshrss instance, or invokes `lab freshrss` / `mcp__lab__freshrss`. Calls the MCP tool first, falls back to the CLI if MCP is unavailable.

Use when a user asks to debug or fix failing GitHub PR checks / GitHub Actions runs (e.g., 'CI is red', 'why is this PR failing', 'fix the failing checks', 'PR checks broken'). Inspects checks via gh, summarizes failure context, and implements when the user has asked for a fix.

Use when addressing GitHub pull request review comments systematically with mandatory resolution tracking. Triggers on "address PR comments", "fix review feedback", "handle PR review", "resolve PR threads", "respond to review", "work through comments", "mark threads resolved", or any mention of systematically handling GitHub PR review feedback. Fetches comments via gh CLI, creates beads for each thread, presents a grouped summary, links commits to review threads, closes beads on resolution, and blocks completion if unresolved threads remain. Use proactively whenever the user is working through PR feedback — even if they don't explicitly ask for the full workflow.

Lab's Glances integration — Cross-platform system monitoring dashboard. Use when the user wants to manage their Glances instance, or invokes `lab glances` / `mcp__lab__glances`. Calls the MCP tool first, falls back to the CLI if MCP is unavailable.

This skill should be used when the user asks to "automate gmail", "export calendar events", "google drive cli", "google workspace admin", "search gmail from terminal", "upload to google drive", "read google sheets cli", "manage google workspace users", "gogcli", or wants to script Google Workspace services from the command line. Do NOT trigger for "GOG", "GOG.com", "Good Old Games", "game downloads", or anything related to the GOG video game platform.

This skill should be used when the user asks to "send notification", "notify me when done", "push notification", "alert me", "Gotify notification", "notify on completion", "send push alert", "get messages", "list applications", "gotify health", or mentions push notifications, task alerts, or Gotify. This skill is also automatically invoked without user request for long-running tasks >5 minutes, plan completion, user input required, or task transitions.

Load the most recent save-to-md session log into a fresh session and brief the new agent on where the prior session left off. Use at the start of a new conversation when the user says "hand off", "pick up where we left off", "resume the last session", "continue from yesterday", "load the last session log", or otherwise signals they want the prior session's context restored. Pair with save-to-md — that skill writes the log; this one reads it back in.

This skill should be used whenever a prompt mentions any of Jacob's named homelab hosts — **tootie, dookie, shart, squirts, steamy, steamy-wsl, vivobook, vivobook-wsl** — or the **WillyNet** LAN. It provides the authoritative map of which host runs which service, IPs, ports, SSH details, storage layout, backup chains, MCP server locations, and known issues. Example triggers: "what's on dookie", "is plex running on tootie", "ssh into squirts", "where's the qdrant container", "which box has the GPU", "check shart's backup status", "the steamy box", "ssh steamy-wsl", "vivobook-wsl logs". Does NOT fire on generic "my homelab" / "my server" / "my NAS" prompts that don't name a specific host. Runtime inventory is generated at `~/.homelab/homelab.md` and `~/.homelab/homelab.json`; `references/homelab.md` is only the report template.

Lab's Immich integration — Self-hosted photo and video management. Use when the user wants to manage their Immich instance, or invokes `lab immich` / `mcp__lab__immich`. Calls the MCP tool first, falls back to the CLI if MCP is unavailable.

Use when working with Jellyfin media server management, libraries, users, playback, metadata, transcoding, server health, or homelab media workflows.

Compose and Compose Multiplatform expert for UI development across Android, Desktop, iOS, and Web. Use whenever the user mentions Compose APIs (@Composable, remember, LaunchedEffect, NavHost, MaterialTheme, LazyColumn, Modifier, recomposition), Compose Multiplatform (commonMain, expect/actual, Res.*, ComposeUIViewController, UIKitView, ComposeViewport), Android TV (tv-material, D-pad, focus, Carousel), Material 3 motion, atomic design systems, design-to-code workflows, Paging 3, or navigation. Activates Review Mode on GitHub PR URLs and review phrases ("review this PR", "what's wrong with this"). Auto-detects Compose projects on session_start. Backed by actual androidx/androidx and JetBrains/compose-multiplatform-core source receipts. See "## When this skill applies" in SKILL.md for the full trigger surface.

Manage bookmarks with Linkding. Use when the user asks to "save a bookmark", "add link", "search bookmarks", "list my bookmarks", "find saved links", "tag a bookmark", "archive bookmark", "check if URL is saved", "list tags", "create bundle", or mentions Linkding bookmark management.

Lab's Loggifly integration — Docker container log alerting via keyword/regex patterns. Use when the user wants to manage their Loggifly instance, or invokes `lab loggifly` / `mcp__lab__loggifly`. Calls the MCP tool first, falls back to the CLI if MCP is unavailable.