Menu
Register and connect to your machines over Tailscale SSH with agents-cli. Use this skill to sync devices from the tailnet, list them, open a shell on another machine, or see agent sessions running across your whole fleet.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Gotchas for the agents-cli headless release. The macOS home base signs, notarizes, and npm-publishes over SSH with no GUI and no Touch ID — three things must be primed or the release stalls. Triggers on: release stuck at sign/publish, errSecInternalComponent, "no npmjs.com bundle", "Missing embedded.provisionprofile", making the release autonomous.
Search, browse, and read agent conversation transcripts across Claude, Codex, Gemini, and OpenCode. Use this skill to find previous sessions, recover context, or inspect what agents have done.
Organize AI coding agents into teams that collaborate on a shared task. Create teams, add teammates, start them, monitor progress, and collect results. Use this skill when you need parallel agent execution. For single-agent dispatch, use `agents run` instead.
Use FIRST for any question about THIS codebase's structure, symbols, or relationships — where a function/class/file is defined, what calls or imports it, what breaks if you change it, how two parts connect, or to load focused source context before editing or reviewing. Routes to the cgraph MCP tools (graph_query / graph_explain / graph_impact / graph_path / graph_context), which serve ranked file:line results, node neighborhoods, transitive blast radius, and token-budgeted source bundles from a resident per-project graph daemon in ~10ms — far cheaper than grepping and reading files. Prefer over blind grep/read for code navigation, dependency tracing, and impact analysis.
Schedule agents to run on a cron schedule or one-shot at a specific time. The scheduler auto-starts on first add. Triggers on: 'schedule an agent', 'recurring job', 'cron', 'daily', 'every weekday', 'run at 2pm'.
Execute a single agent headlessly or interactively. Supports plan/edit/auto/skip modes, secrets bundle injection, version pinning, fallback chains, balanced rotation, profile dispatch (Kimi/DeepSeek/etc.), and workflow dispatch by name. Triggers on: 'run claude', 'run codex', 'agents run', 'dispatch an agent', 'headless agent', 'one-off agent task'.
| name | devices |
| description | Register and connect to your machines over Tailscale SSH with agents-cli. Use this skill to sync devices from the tailnet, list them, open a shell on another machine, or see agent sessions running across your whole fleet. |
| argument-hint | [sync|list|show|add|set|ssh|update|run|fleet] |
| allowed-tools | Bash(agents devices*), Bash(agents fleet*), Bash(agents ssh*), Bash(agents sessions*), Bash(agents run*), Bash(agents hosts*), Bash(agents logs*) |
| user-invocable | true |
Manage a registry of SSH device profiles and reach your other machines. The
registry self-populates from tailscale status --json, so on a tailnet you
rarely hand-enter a host. This skill teaches the agents devices /
agents fleet (alias) and agents ssh CLIs, plus the fleet-wide
agents sessions --active view.
agents fleet is a synonym for agents devices — every subcommand works under
either name (agents fleet list == agents devices list).
# Ingest the tailnet into device profiles. In a terminal this opens a
# checkbox to register/unregister nodes; --yes registers every non-ignored one.
agents devices sync
agents devices sync --yes
# Register or dismiss a single discovered node.
agents devices register <name>
agents devices ignore <name> # never suggest it again
agents devices unignore <name> # undo an ignore
agents devices list # platform, address, reachability (alias: ls)
agents devices list --json # registry as a JSON array (for scripts/hooks)
agents devices show <name> # full profile for one device
list marks the machine you are on with ▸ <name> ← this machine.
# target is user@host or host; platform is windows | linux | macos
agents devices add <name> <target> --platform linux
# Update fields on an existing device.
agents devices set <name> --user muqsit --platform macos
agents devices set <name> --auth password --bundle <secrets-bundle>
agents devices rm <name> # remove from the registry (alias: remove)
Auth is either key (system ssh agent / on-disk keys) or password (pulled
from a Keychain-backed secrets bundle — never stored on disk).
# Open a shell (preflights reachability, picks the shell, authenticates).
agents ssh <name>
# Run a one-off command and return.
agents ssh <name> uname -a
# Roll out the latest agents-cli to every online registered device.
agents fleet update
agents fleet update 1.20.62 # pin a version / dist-tag
# Run an arbitrary command on every online device; offline ones are skipped.
agents fleet run uname -a
agents fleet run 'agents --version'
Both print a per-device result table (ok / failed / skipped).
To use plain ssh <name>, render the registry into your ssh config:
agents devices render # print the Host stanzas to stdout
agents devices render --write # write ~/.ssh/config.d/agents
agents sessions --active groups running agent sessions by machine, pins the
local box first (▸ <name> ← this machine), and folds in sessions from your
registered, online devices over SSH:
agents sessions --active # this machine + every online device
agents sessions --active --local # this machine only (no SSH fan-out)
agents sessions --active --host zion --host mac-mini # specific machines
agents sessions --active --json # merged, machine-tagged, for scripts
Unreachable or CLI-less hosts are skipped with a note, never fatal. If no
devices are registered, it prints a tip pointing you at agents devices sync.
Offload agents run itself to a registered host over SSH — it follows live by
default; --no-follow detaches and returns immediately.
agents run claude "profile this build" --host gpu-box # run there, follow live
agents run claude "..." --host gpu-box --no-follow # detach
agents hosts ps # list dispatched runs and their status
agents hosts stop <id> # terminate a hung/detached run (alias: kill)
agents logs --host gpu-box # pick a run on that host and view its log
agents logs <id> -f # re-attach to a running one and follow
agents logs [id] is the unified viewer — it resolves a host-dispatch run OR a
local session by id, filters with --host/--agent/--version, and -f
follows a live one. agents hosts logs <id> is the host-only equivalent.
agents hosts stop <id> SIGTERMs the remote process group from this machine and
keeps the log for later inspection.
Registered devices are part of the host pool automatically: they appear in
agents hosts list (SOURCE devices), resolve as --host targets by name,
and join capability routing once tagged (agents hosts add <device> --cap gpu
enrolls straight from the device profile — no target needed). Password-auth
devices are listed but can't carry a dispatch (BatchMode ssh); switch them with
agents devices set <name> --auth key. Devices also work as cloud task targets
(agents cloud run "…" --host <device>) and routines placement
(agents routines add … --run-on <device>).
list is a snapshot from the last sync; rerun sync to
refresh it.--active fan-out.agents devices set <name> --auth password --bundle <name>.