Skip to content

The automations

Seven projects live in the repo. One is the front door; the other six are the agents that do the work. As the vision lays out, every agent is either a writer (structures messy input and stores it in Attio) or a retriever (pulls knowledge back out, read-only, when it's most useful).

Automation Kind What it owns
u2v-dispatch Front door Owns the inbox, classifies each email, routes it to a workflow, replies
u2v-dealflow ✍️ Writer A startup (deck / link / name) → scored, classified, written to Attio
u2v-lp-intake ✍️ Writer An LP lead (website / roster) → researched, deduped, written to Attio
u2v-lp-tasks ✍️ Writer A one-line request → a follow-up task created on an LP in Attio
u2v-lp-share 🔎 Retriever An LP → an Excel snapshot + a branded "what's new" briefing
u2v-lp-match 🔎 Retriever An LP or sector → a ranked shortlist of dealflow to show them
u2v-dealflow-sample 🔎 Retriever (autonomous) A refreshed sample deck, emailed to the team bi-monthly

How to read these pages

Each automation page follows the same shape, so once you've read one you can skim the rest:

  1. What it does — the job in a sentence or two.
  2. The workflow — the end-to-end path from input to result, as a diagram.
  3. Key decisions — the choices that shaped it, and why they were made that way.
  4. Where things live — the files and folders that matter, so you can find them in the repo.
  5. Status & how to run it.

The pages are a map, not the territory

They're written to get you oriented fast — enough to understand the architecture and find the right file. When you need line-level detail, open the repo in Claude Code and ask about the specific piece; each project's CLAUDE.md gives it the context to answer well.

Start with u2v-dispatch — it's the front door every routed workflow passes through.