Companies starting point

Thank you Aron! Amazing feedback and super cool to see how you think. I will start playing with this.

Nice, Niche Site Empire is a good one to start with because the org chart and the first tasks are already wired to a clear opening move — you don't actually have a hero site yet, so half the roster is idle until that gets resolved. That's your starting point.

Here's the concrete sequence I'd give the CEO on first wake:

Phase 0 — pick a niche before anything else ships. The org has 16 agents, 2 projects, and 6 starter tasks, but five of those six tasks are "…hero site…" and there is no hero site yet. So nothing technical, affiliate, or schema-related can run until the CEO has a domain to point them at. First real task to assign:

> "Niche Researcher: propose 3 keyword-cluster candidates that meet our 80/20 economics rule. For each: estimated traffic ceiling, monetization fit (Amazon/ShareASale vs. display-ad heavy), competitive density, and a kill threshold. Return to me for approval before any domain diligence."

This is keyword-cluster-research doing what it's there for, and it forces the CEO to actually commit to a niche instead of letting the whole portfolio sit in superposition.

Phase 1 — domain, once the niche is approved. Hand diligence-and-acquire-hero-domain to the Domain Acquirer with a hard constraint: Wayback-clean, no manual actions, niche fit. The skill domain-acquisition-diligence already encodes the checklist. Don't let it skip to a purchase without your sign-off — this is an irreversible spend.

Phase 2 — the four parallel tracks unlock. Once the domain exists, the rest of the starter tasks become real work and can run in parallel: deploy-hero-site-schema-and-cwv-baseline → Technical SEO Lead score-hero-cluster → Niche Researcher (now scoped to the chosen niche) onboard-hero-site-affiliate-networks → Affiliate Program Manager ship-first-40-hero-cluster-articles → Content Director (gated on schema + cluster)

The prepare-mediavine-application-package task is a trap to enable too early — Mediavine wants 50K sessions/month, which won't happen for months. Park it; don't let the Display Ads Manager burn cycles on it yet.

Two things to watch specifically for this template: Heartbeats stay off on the publishing-side agents (Content Director, Editor, Writers Lead) until you actually have briefs flowing — otherwise they'll wake up every interval and report "nothing to do." Only the CEO and the Niche Researcher need to be reachable in Phase 0. The kill rule is load-bearing. The identity says "any site under $200/month after 9 months goes on the sunset list" — make sure that constraint is in your CEO's first heartbeat context, otherwise it'll quietly defend losers six months from now.

When the CEO comes back with the niche shortlist, that's your real first decision point. Everything downstream is mechanical once you've committed to a niche.

Thank you! Will go through your answer. I downloaded the Site Niche Empire template.

Solid feedback — you're right, this is a gap. The community templates spec ([paperclipai/companies](https://github.com/paperclipai/companies)) already recommends shipping a README.md with a getting-started section, but two things are missing: it's a convention rather than a spec requirement, and even when a README exists, Paperclip doesn't surface it anywhere after import. I'll add both to the roadmap for future implementation — making the README a required field on the agentcompanies/v1 schema, and showing it in the UI on the first visit to an imported company. In the meantime, here's the generic post-import checklist that applies to any imported company:

Review what was imported, agent by agent. Open Org and click each agent. Verify the adapter type and working directory look right for your machine — paths and adapter choices that worked on the exporter's setup may not match yours. Add the secrets the package declared. Exports never include API keys — secrets are environment-specific. The .paperclip.yaml in the package lists which env vars the agents expect; fill those in via instance/company secrets before enabling anyone. Without these, the first heartbeat will fail at adapter startup. Set per-agent budgets. Imported agents inherit the exporter's budget values, which may or may not match what you're comfortable spending. Even $20/agent/month is a useful guardrail to start. Read the company goal and tweak it. A template's goal is generic ("Run a content engine"); make it yours ("Build a content engine for [your product] targeting [your audience]"). The CEO reads this every heartbeat, so any vagueness here propagates downstream. Enable heartbeats deliberately, not all at once. Imported agents always start with timer heartbeats off — that's intentional, not a bug. Most agents stay healthier on "wake on demand" only; reserve timer heartbeats for agents that genuinely need to poll something external. Background: Why Agents Do Nothing by Default (/guides/concepts/why-agents-do-nothing-by-default). Hand the CEO its first real task. Even with a great template, the CEO won't run until something wakes it. Assign a concrete first task — "Read the goal, audit the agent roster, propose the next 3 tasks to me for approval" — and it'll get going. The seven questions in Before You Hire Your First Agent (/guides/getting-started/before-you-hire-your-first-agent) apply just as much to inherited agents as to ones you build from scratch.

If you tell me which template you downloaded, I can probably point at more specific first-task ideas for that org structure too. And if the template you used did include a README that you missed, that's exactly the surfacing-in-UI problem above — worth knowing.