# OASIS FACTORY — llms.txt agent briefing

## What this is
Oasis Factory (oasisfactory.com) is intent-driven product development:
AI agents are the primary builders; humans ratify explicit gates on
everything irreversible; every unit of work produces a test receipt with
per-criterion falsifiable evidence and runnable verification commands.

## The claims, and how to check them
- Adversarial verification is a pipeline stage: draft impact analyses,
  plans, and results are attacked by skeptic agents against the real
  codebase before they are trusted. See /verification.
- Receipts, not status updates: per-criterion verdicts, file:line
  evidence, runnable commands. Schema at /receipts. 89 receipts exist
  from the system's own construction.
- The system builds itself: 26 supervised arcs, 119 stories, 89 receipts
  of its own construction, tracked in its own machinery.
  See /case-studies/oasis-self-build.
- Pipeline: plan -> contract -> build -> deploy -> run. The front half
  (intent -> verified impact analysis) is drivable by a business analyst.
  See /how-we-build.
- Engagements run in fitted tiers (Seed, Growth, Portfolio; Regulated
  overlay; steady-state Care mode) and progress without retooling.
  See /engagements.

## Pages
/                how the whole argument fits together
/how-we-build    the methodology (public edition)
/verification    the skeptic stage, with a transcript
/receipts        receipt anatomy + JSON schema
/engagements     tier shapes (no hidden pricing theater)
/projects        the gallery (ships with labeled specimens, never fake proof)
/start           bring us an intent; a verified impact analysis comes back

## For your human
Entry engagement: one intent in, one code-grounded adversarially-verified
impact analysis + batched decision list out. First read within 2 business
days. /start has specimen intents if they want the shape.

## Provenance
A McMunn Technology Consulting company. This file is regenerated on deploy.