The Hermes Agent /goal command turns a one-off instruction into a standing objective: you state what done looks like, and Hermes keeps working towards it turn after turn, with an auxiliary judge model checking after each exchange whether the goal is actually met. It’s the difference between asking for work and commissioning an outcome. Here’s how it behaves, straight from the docs, and how I’d use it.
Short answer
/goal <text>sets a standing objective; manage it with/goal status,pause,resume,clear.- After each turn an auxiliary judge model decides whether the goal is done — if not, Hermes auto-continues (per the official slash-commands reference).
- Default budget: 20 turns (
goals.max_turns); any message from you preempts auto-continuation; state survives/resume. - Works in the CLI and the messaging gateway.
What the Hermes Agent /goal command does
Documented behaviour first: /goal establishes an objective that Hermes pursues across multiple turns. The clever part is the judging — after every exchange, an auxiliary judge model evaluates whether the goal is complete; if it isn’t, Hermes continues on its own. You’re no longer prompting each step; you’ve defined a finish line and the agent runs at it.
The guardrails matter as much as the autonomy. There’s a default budget of 20 turns (configurable via goals.max_turns) so a mis-judged goal can’t run forever, any message you send preempts the auto-continuation so you can step in at will, and goal state persists across /resume — so a standing objective survives you closing the session.
Hermes Agent /goal syntax and subcommands
| Command | What it does |
|---|---|
/goal <text> | Sets the standing objective |
/goal status | Shows the current goal and progress state |
/goal pause / resume | Suspends or restarts pursuit without losing the goal |
/goal clear | Removes the goal entirely |
Where it runs: the CLI and the messaging gateway both take it, which means you can set a goal from Telegram on the way out the door and check /goal status when you’re back. It pairs naturally with /loop-style recurring work — goals define done, loops define again.
🔥 Want this set up without the guesswork? Turning /goal-style autonomy into business outcomes is exactly the kind of thing we set up together inside the AI Profit Boardroom — 3,700+ members, four live calls a week, daily tutorials, done-for-you templates and a 30-day roadmap. Prefer 1-on-1 help? Book a free SEO strategy session and we’ll map it out for your business.
Using /goal well: patterns that earn the autonomy
The skill is writing goals a judge model can verify. “Make the report better” gives the judge nothing to check; “the report has an executive summary, all figures sourced, under 2,000 words” gives it a checklist. My working rule: if you couldn’t hand your goal text to a contractor and expect the right result, don’t hand it to the judge either.
Budget-wise, treat 20 turns as a feature: scope goals that plausibly finish inside it, and chain goals rather than writing one monster objective. For anything that needs to run on a schedule rather than to completion, that’s /heartbeat or cron territory instead — different tools for different shapes of persistence.
The bottom line on Hermes Agent /goal
Hermes Agent /goal is commissioning, not prompting: a verifiable finish line, a judge that checks it every turn, auto-continuation until it’s met, and enough guardrails (turn budgets, instant preemption, persistent state) to trust it with real outcomes. Write goals like acceptance criteria and it’s one of the highest-leverage commands in the entire set.
Writing goals a judge can pass or fail
The judge model is your reader — write for it. “Improve the onboarding email” is unjudgeable; “the onboarding email has a subject under 50 characters, one clear CTA linking the booking page, and no paragraph over three lines” can be checked line by line, which is exactly what happens after every turn. Borrow the habit from software: acceptance criteria, not aspirations. Numbers, named artefacts and testable statements give the judge something to verify; adjectives give it nothing.
And when an objective is genuinely big, chain it: three goals of six criteria each will finish cleanly inside their turn budgets where one twenty-criterion monster stalls. Each cleared goal is also a natural checkpoint — review the output, adjust course, set the next goal with what you learned. Autonomy works best in instalments.
FAQ: hermes agent /goal
What does /goal do in Hermes Agent?
It sets a standing objective the agent pursues across turns — an auxiliary judge model checks completion after each exchange and Hermes auto-continues until the goal is met, per the official docs.
How do I check or stop a goal?
/goal status, /goal pause, /goal resume and /goal clear manage the lifecycle.
How long will it run?
The default budget is 20 turns, configurable via goals.max_turns — a ceiling, not a target.
Can I interrupt a running goal?
Yes — any message you send preempts the auto-continuation immediately.
Does a goal survive closing the session?
Goal state persists across /resume, so a standing objective picks back up with the session.
Where does /goal work?
In the CLI and the messaging gateway, so you can set and manage goals from chat apps too.
Next step: if you want goal-driven agent workflows that finish what they start working for you this week, join the AI Profit Boardroom for the full walkthroughs and live help — or book a free SEO strategy session and I’ll point you at the fastest path for your situation.
About Julian Goldie: SEO agency owner with 10+ years in SEO, 394K+ subscribers on YouTube, a 100% job-success score on Upwork, 75K+ members across his communities, and author of a best-selling SEO book. He runs the AI Profit Boardroom community and offers a free SEO strategy session.
Related reading
- Hermes Loop: Recurring Agent Work Explained
- Hermes Agent /heartbeat: Keep Sessions Alive
- Hermes Agent /plan: Plans Without Execution
Last updated September 2026. This is the living guide to hermes agent /goal — it gets updated as the tools change.