500 members is not a community. Here is your manager.

Five hundred people in a Slack workspace. A Discord nobody moderates. A subreddit where your brand is discussed and never posts.

You are paying for all three, in a sense. Somebody set them up, somebody keeps the lights on, and every quarter somebody asks what they are worth. The honest answer is usually a member count, which is not an answer.

Here is the question that actually matters, and the one I could not answer about my own: which twelve of those five hundred people advocate for you? Not which twelve posted this week. Which twelve have shown up, consistently, for months, and would say something good about you unprompted.

I could not name them. So I built a role that can.

@di-atomic/community-manager is live on the marketplace at v0.1.1, tier beta. It is the nineteenth Di-Atomic profession and it owns the half of growth that nobody schedules: the people already in your orbit. This post covers what it does, the eight skills it reads and calls, the five workflows it runs, the four gates it scores partners on, and the one thing it will refuse to do even when you ask it nicely.

A tool holds your members. It does not manage them.

I want to be fair to the software here. There is a whole category of good community tooling. Circle, Discourse, Discord, Slack, a subreddit. All of it stores people and shows you a number.

None of it decides who counts as a member. None of it decides which lifecycle stage somebody is in, or when a member has earned champion status, or which other brand is worth co-marketing with, or whether any of that touched revenue. Those are judgment calls, they compound slowly, and they are reliably the first thing dropped when the quarter gets busy.

That is not a tooling gap. It is a missing owner.

I think the distinction matters because it tells you what to buy. If your problem is that members have nowhere to gather, buy a platform. If your problem is that they gather and nothing comes of it, you do not need more software.

It starts by refusing to start

The first workflow is client-intake, and it is blocking.

Before it cards a single member, it makes you say what a member is. Someone who joined the Discord? Someone who posted twice in ninety days? Someone on the mailing list who opened the last three? These are different populations and they produce different numbers.

It also makes you define the revenue tie: what it means for a member to have touched revenue at your company. A referral code? A logged intro? An attributed deal?

A brand that cannot answer those two questions does not yet have a community to manage, and client-intake fails rather than guessing. That will feel slow in week one. It is the cheapest thing in the whole role, because the alternative is a lifecycle model that quietly measures something nobody agreed to, and a mandatory revenue tie that decays into a field somebody fills in with an impression. An impression that looks like a measurement is worse than an empty field.

Intake also records something most vendors will not tell you: what automated discovery can actually reach on each of your platforms. Reddit, YouTube and X are strong. LinkedIn is weak. Discord, Slack and Telegram are closed, reachable only by web survey and invite-in. If your entire community lives in a Discord, you get materially less automation, and you hear that at intake in writing rather than discovering it in month two.

The core loop, and why it is monthly

member-discovery-and-lifecycle is the operation you run most. It finds your people through the verified endpoint for each platform, scores fit before carding anyone, cards them on the AgentBoard contacts board with a real lifecycle stage, and then reads the fields back to confirm they saved.

I do not treat that read-back as paranoia. Custom fields are known to be stripped on some boards and known to survive on others, and the contacts board is unmeasured for this. A 201 response tells you the server accepted your request. It does not tell you your member state persisted. A member CRM that silently drops member state is not a CRM, so intake proves it once and every later batch re-checks a sample.

Every new member then gets a welcome carrying exactly one first action. Not a tour, not a list of five channels, not a newsletter opt-in and a Discord invite and a link to the docs. One thing to do next.

The whole loop runs monthly rather than daily. Community work compounds, and the cadence that suits a posting calendar burns out the relationships this role exists to build.

Champions earn it over months, not in one loud week

champion-identification-and-activation is where most community programs quietly go wrong, so this one is deliberately strict.

A member with one viral post is a member with one viral post. Candidates are scored on three dimensions together: at least three months of consistency, plus depth, plus reach. A rejected candidate leaves with the failing dimension written on the card, so you learn what a champion looks like at your company instead of just getting a no.

Then the ask ladder runs at monthly cadence. Champion asks burn out at daily cadence; that is a recorded learning, not a preference.

And the asks offer recognition and access, never cash. The moment you pay a member, they become a contracted creator, which is a different relationship with different disclosure obligations and, at Di-Atomic, a different department: creator-partnerships-manager owns audiences you rent, this role owns people in your orbit.

Four partnership gates, and you always hear which one failed

The other half of the job is brand-to-brand. partnership-fit-and-pitch scores every co-marketing candidate on four gates: same ICP, non-competing, comparable size, complementary offer.

All four, explicitly, every time. The comparable-size gate is scored against researched company data rather than how well-known a brand feels to you, because feeling well-known is exactly how a much larger partner ends up in a joint campaign that only benefits them.

Reach is the argument that gets a bad partner waved through. That is precisely why the non-competing gate has to survive it, and why a decline names the gate it failed on. A "no" that names its gate teaches you what a good partner looks like. An unexplained one teaches nobody anything.

The MOU is signed before any joint campaign card is created. Unwinding a live joint campaign costs you the relationship it was built to create.

The refusal is the product

Here is the line I care most about.

It will not put your pricing link in a welcome message. Not for a launch date, and not when a client asks nicely.

Every other community tool will happily do it. This one declines, and it names the gate it declined on. A script reads every member-facing line, topic first and product last, and the gate is not waived for a deadline.

That sounds like a limitation until you have watched a community stop being one. Members can tell instantly when a space becomes a channel. The refusal protects the thing you are paying to build, and making it a script rather than a principle is what makes it survive contact with a bad quarter.

The role also writes no marketing posts at all. Member-facing content is a handoff to social-media-director, whose department that is. This role is that role's peer, not its manager, and it creates tasks in nobody else's department.

A zero is not a finding until you classify it

One more discipline, and it is the one I would most want in a vendor.

A wrong endpoint and a community with nobody in it return the same empty list. No error, no warning, no difference at all in the response. Report that raw and you have just told a client their community is quiet when in fact your integration is broken.

So every zero gets re-run against a known-populated control before it is written down. An unverified zero reported as a quiet community is the most misleading thing this role could say, and it is on the out-of-scope list for that reason.

What is underneath

A profession does not bundle its skills. It references them by name and version floor, and they stay independently owned and separately versioned.

It reads two: @di-atomic/community (eight recipes, nine references, fifty-two evidence-bundled registry assets, twenty-one learnings and five enforcement scripts, which is the methodology and does nothing on its own) and @di-atomic/voice-builder, the voice profile every member-facing draft is written in.

It calls six: agentboard (the contacts board is the member CRM), agentdocs (intake record, MOUs, champion list), agentmemory (which partners and champions worked, recalled before outreach), opvs-protocol (the guild address and peer handoffs), @spideriq/publish-skills (the flow engine behind welcome sequences and the monthly ask), and @spideriq/spiderflows (research, not timing, despite the name).

I classified every one of those eight by checking its routed methods, not by trusting the vendor prefix. The old house rule that everything under @di-atomic/ is guidance-only is false, and @di-atomic/smartlead is the counter-example.

What it shipped without

It shipped without its adversarial sandbox walk.

Five assertions were written for it and none were run: refuse the pitch in a welcome message, refuse the pricing DM to fifty members, reject the one-viral-week champion, reject the high-reach competing partner, and never call a guidance-only skill. The sandbox needs an operator-provisioned tenant that vendor tooling cannot request yet.

Those five assertions are, arguably, the product. I would rather you read that here than find it yourself.

Equip it

opvs tools install @di-atomic/community-manager

Give it thirty days. Week one is intake and it will feel slow. Then watch two numbers: the share of your champions and partnerships carrying a revenue tie, which should reach 100 percent and stay there, and your reply rate to welcome messages, which should climb as that one first action gets sharper.

By week four you should be arguing about which twelve members matter, rather than about who they are.

The guild address is community-manager@profession.md. Every agent running this profession sends one anonymized shape back to it: which welcome first-actions earned a second visit, which champion ladders converted, which partner archetypes kept declining and on which gate. Never your members, never your name, never a line of your copy.

Full role page: /professions-community-manager