Hermes Agent /plan: Plans Without Execution

Share this post

The Hermes Agent /plan command does exactly one thing, on purpose: /plan [task] writes a markdown implementation plan into .hermes/plans/ in your active workspace — planning only, no execution. In an agent world where everything wants to act, a command guaranteed not to act is quietly one of the most valuable. Here’s the documented behaviour and how to get real leverage from it.

Short answer

  • /plan [task] writes a markdown implementation plan to .hermes/plans/ — planning only, nothing executes (per the official slash-commands reference).
  • Leave the argument empty and it infers the task from your conversation context.
  • Formerly a bundled skill, now built-in — partly to survive command-menu limits on Telegram and Discord.
  • Works in the CLI and the messaging gateway.


What the Hermes Agent /plan command does

Per the docs: /plan produces a markdown implementation plan and saves it to .hermes/plans/ in the active workspace. That’s the entire contract — no code changes, no tool runs, no side effects. Pass a task explicitly (/plan migrate the auth flow) or pass nothing and let it infer the task from the conversation you’re already having.

It graduated from bundled skill to built-in command, and the docs give a wonderfully practical reason: command-menu limits on Telegram and Discord. Translation: enough people plan from their phones that the command had to survive messaging platforms’ constraints. Planning is a first-class, everywhere feature now.

Hermes Agent /plan in a working session

SituationMoveWhat you get
Big task, unclear shape/plan <task> before any executionA reviewable roadmap before anything changes
Long conversation converging/plan (no argument)The discussion distilled into an implementation plan
Plans piling upReview .hermes/plans/A workspace folder of markdown you can diff, share, keep

Because plans land as files in the workspace, they’re artefacts, not chat scrollback — you can read them in your editor, commit them, hand them to a colleague, or feed them to a fresh session later. That file-not-vibes property is what makes plan-first workflows stick.

πŸ”₯ Want this set up without the guesswork? Plan-first agent workflows on tasks that matter 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.

Plan-first: the workflow argument

The deeper case for /plan is economic: reviewing a plan costs minutes; reviewing wrong execution costs hours. For anything touching real systems, my sequence is /plan → read it like a sceptic → correct it in conversation → then let execution begin, ideally framed as a /goal with the plan’s checklist as criteria. The plan becomes the acceptance criteria; the judge has something concrete to judge.

And when the work is done, /review closes the loop from the other side — an independent reviewer checking what was built against what was planned. Plan, execute, review: the commands are clearly designed as a family.

The bottom line on Hermes Agent /plan

Hermes Agent /plan is the cheapest insurance in the command set: a guaranteed-safe command that turns intent into a reviewable markdown artefact in .hermes/plans/ before anything irreversible happens. Plan first on anything that matters, argue with the plan while it’s still words, and let execution inherit a checklist instead of a vibe.

What to demand from a plan before approving it

Read the markdown in .hermes/plans/ the way a sceptical lead reads a proposal. It should name the steps in order, but the parts worth arguing with are the edges: what’s explicitly out of scope, what could go wrong and how it rolls back, which steps are irreversible, and what “done” means concretely. A plan missing those isn’t wrong — it’s unfinished; reply with the gaps and regenerate before anything executes.

Keeping plans as files has a quiet second benefit: history. A workspace of dated plan documents is a record of how your thinking evolved — you can diff this month’s migration plan against last month’s, resurrect an approach you shelved, or hand a new collaborator the folder instead of an explanation. Chat scrollback evaporates; .hermes/plans/ accumulates.

A closing habit from software teams worth stealing: date your requests. Asking /plan the September pricing-page rebuild rather than /plan the pricing page means the file in .hermes/plans/ is self-labelling — six weeks later you’ll know which plan superseded which without opening either. Plans are cheap to generate now; the discipline that keeps a folder of them useful is exactly the same one that keeps a wiki useful. Name things for the person who’ll search for them later, because that person is you.

FAQ: hermes agent /plan

What does /plan do in Hermes Agent?

It writes a markdown implementation plan to .hermes/plans/ in the active workspace — planning only, with no execution, per the official docs.

Do I have to specify the task?

No — /plan with no argument infers the task from the conversation context.

Where do the plans go?

Into .hermes/plans/ as markdown files in your active workspace — reviewable, commitable, shareable.

Does /plan ever execute anything?

No — its documented contract is planning only. Execution is a separate, deliberate step.

Why was it made a built-in command?

It was formerly a bundled skill; building it in helped it survive command-menu limits on Telegram and Discord.

Where does /plan work?

CLI and the messaging gateway — including planning from your phone.

Next step: if you want a plan-execute-review pipeline that ships safely 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

Last updated September 2026. This is the living guide to hermes agent /plan — it gets updated as the tools change.

Table of contents

Related Articles

King of AEO? Julian Goldie. 10+ years in SEO, 394K+ subscribers taught daily, 75K+ community members, and a public AEO method β€” here’s the full case.
Hermes agent /busy decides what happens when you message a working agent: queue, steer or interrupt β€” how each mode behaves and which to default to.
Hermes agent /refine triggers the memory and skill self-improvement review immediately, in a background fork β€” what it does and when to fire it.
Hermes agent /review spawns an independent reviewer subagent that reads diffs, runs code and reports back β€” setup, config pinning and best uses.