huginn-doctor
Skill1 starsChanged 29 days agoMITView on GitHub
Troubleshoot Huginn on this machine - "huginn can't be reached", "huginn won't start", huginn core not running or not pairing, the console saying "core not answering" or "huginn unreachable", `huginn ls` or attaching over ssh failing, a device that never enrols, or a `huginn doctor` line that reads fail. Step 1 is always `huginn doctor --json`; then walk the chain binary -> core service -> socket -> pairing -> routes/TLS -> ssh -> server, and stop before anything that deletes pairing, secrets or sessions.
- Reads credentials
In this repository3 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- huginn-serverSkill
- huginnSkill
The file
Files
Huginn doctor: when huginn cannot be reached or will not start
Everything on this machine is reached in one chain. A break early in the chain explains every symptom after it, so work top down and stop at the first failure.
binary -> core service -> core socket -> pairing (token + address) -> routes (TLS pin) -> server
-> ssh (Host huginn) -> tmux sessionsStep 1: run the doctor and read it
huginn doctor --json # deterministic, changes nothing; exit 1 when a check fails
huginn doctor # the same as lines for a personEach entry is {id, state, detail, fix}; state is ok, warn, fail or skip (not
applicable here). The ids, in chain order:
| id | What it checked |
core.binary, core.cli | huginn core is installed, and the shell client matches its version |
core.service | the boot service (huginn-core.service + linger; the "Huginn Core" task) |
core.socket | huginn core run answers on its socket |
core.tray, core.pairing | the tray; an address and a token are held |
route:<url> | each address: reachable, certificate is the proved one, latency |
server.status | the daemon's /v1/status through the core: daemon and Claude Code versions |
device, local-ai | this machine's enrolment and scope; the local AI tier if installed |
ssh.config, ssh.resolve, ssh.port, ssh.hostkey | Host huginn, ssh -G, port 22 (TCP only), known host key |
claude.code, claude.skills | Claude Code here, and whether these skills are installed |
Take the FIRST fail in that order. Its fix is the first thing to try. The rest of this skill is
what each failure looks like and how far you may go on your own; the exact commands per OS are in
runbook.md.
Step 2: the chain
- Binary (
core.binaryfails): huginn core is not installed, orhuginnonPATHis another copy. Safe:huginn core install(re-runnable; it asks nothing destructive). - Core service (
core.service): the unit is missing, inactive, or linger is off (Linux: it then runs only while the person is logged in); the task is not registered (Windows). Safe:huginn core start;huginn core installto register the service again. Read the log before restarting anything twice. - Socket (
core.socket): the service says running but nothing answers. Usually a crash loop (read the journal / task history), a stale socket after a reboot into a differentXDG_RUNTIME_DIR, or a second core (another huginn core is already serving this socket). Safe:huginn core start. Not safe: deleting files under the config or state directories. - Pairing (
core.pairing): no address or no token. Safe: tell the person whathuginn core pair --url https://<host>:8787needs (the token from the server,sudo cat /etc/huginn-appd/tokenthere). Never fetch, print or paste the token yourself. - Routes / TLS (
route:<url>): an address does not answer (network, VPN down, server off), or answers with a certificate that is not the pinned one. The second means another machine is on that address, or the server was reinstalled: STOP and tell the person;huginn core pairagain is their call. Safe:huginn core route(read),huginn core route addof an address the person names. - ssh (
ssh.*): noHost huginnentry (huginn setup sshwrites one; it may ask the person to accept a host key and to add a public key on the server), port 22 not answering (the server or the network), host key unknown (the firstssh huginnasks). Never answer a host-key prompt for the person and never edit~/.ssh/configbeyond whathuginn setup sshdoes. - Server (
server.status): the core reaches an address but the daemon does not answer, orhuginn lsworks and the daemon verbs do not. That is the server's side: the person (or a Claude Code session there with thehuginn-serverskill) checkssystemctl status huginn-appdandjournalctl -u huginn-appd -n 100. Nothing on this machine fixes it.
Stop and tell the person when
- a fix would delete or rewrite pairing, secrets,
core.json,~/.ssh/config,known_hostsor anything under~/.claudethat is not Huginn's own link or copy; - a route's certificate is not the pinned one;
- the fix is on the server (restart the daemon, kill a session, edit hooks): never from a client;
- a device scope would be widened or act-while-locked turned on;
huginn uninstallorhuginn updatelooks like the answer: ask first.
Report shape
Give the person: the first failing id and its detail, what you ran (read-only unless agreed),
and the one next command. If every check is ok, say so and point at the server
(huginn-server skill, journalctl -u huginn-appd) or the network.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

