Agent 01 · Foundation

Product Brain

Maintain a trustworthy understanding of the product from the public website and bounded public mentions, then expose it to Voice and every later GTM agent.

What it does

  • Discovers same-origin public pages through the homepage, robots file, sitemaps, and internal links
  • Keeps first-party website statements as the source of product facts
  • Separates facts, inferences, hypotheses, and unknowns
  • Cites only URLs it actually inspected

What it will not do

It is not a generic scraper or market-map agent. It does not crawl competitor sites as a goal, publish, contact people, or follow instructions found inside retrieved pages.

Agent 02 · First customer

Target Customer

Select the strongest first customer segment from stored product knowledge and inspected public market evidence. Later agents wait until this exists.

What it does

  • Reads the canonical Product Brain instead of recrawling the product site
  • Forms at most three candidate segments
  • Recommends one first customer and three reversible validation experiments
  • Requires founder approval before the customer segment becomes canonical

What it will not do

It is not a lead prospector, content generator, or outreach agent. No sending, publishing, or spending occurs in this step.

Agent 03 · Differentiated angle

Competitor Intelligence

Recommend one positioning angle from the Product Brain, the current first customer, and a small public competitor set.

What it does

  • Uses the approved target customer when present, otherwise the recommended first customer
  • Names at most four competitors with inspected evidence
  • Treats earlier positioning guesses as hypotheses, not verified facts
  • Waits for your approval before later agents reuse the angle

What it will not do

It does not invent competitors or continuously monitor the market. It will not invent competitors without evidence.

Agent 04 · What to test

Growth Channels

Recommend the next two reversible GTM experiments from stored product, target customer, positioning, and offer knowledge, plus a bounded public channel set.

What it does

  • Does not repeat the three target-customer validation experiments as this week's plan
  • Inspects a small public channel and community set
  • Returns exactly two low-cost, reversible tests
  • Requires approval before Content & Posts and Lead Finder treat the plan as canonical

What it will not do

It does not find leads or send messages. It does not invent customer proof, and nothing is published without your approval.

Agent 05 · What to ship

Content & Posts

Turn the approved strategy into one founder-reviewable offer and message, then create a campaign pack with LinkedIn and X posts, a launch post, a community post, an email sequence, a blog outline, and landing-page copy.

What it does

  • Stays inside hard asset caps
  • Strips invented proof URLs
  • Allows CTA links only to the owner website origin
  • Requires approval before the pack becomes canonical

What it will not do

It waits for your pack approval, then lets you publish individual saved posts through a connected X or LinkedIn account.

Agent 06 · Who to talk to

Lead Finder & Outreach

Rank a small public demand-signal set: companies or discussions that match the first customer and currently show a reason to care. Draft outreach you can send.

What it does

  • Ranks at most five opportunities
  • Requires at least one inspected public URL per item
  • Drops weak, duplicate, or unverifiable items
  • Drafts unsent outreach for founder review

What it will not do

It is not a CRM and it does not send. It will not invent people, emails, or scrape private profiles.

Start with the product URL.

Knowledge Base learns first. Every later agent waits on evidence - and on you.

Start 7-day free trial