History in, offer sheet out

Point it at the traces your work already leaves — AI session history first, then chat, tickets, or files if you allow them. Worker agents normalize those traces into one SQLite database (corpus.db), fresh auditors verify the rows against their sources, and you get an offer sheet of two to five automations, each backed by a query you can run against your own history.

You pick one, or none. It builds only what you chose, then proves the build by executing the receipts the offer promised. It never decides for you.

The bar is strict on purpose: a candidate needs at least three independent occurrences in your history. One busy thread is one event, not three.

“Nothing worth building” is a real answer

Before reading anything, it asks what is already automated — by you or your teammates — because automations running under a human account look exactly like hand-work in the traces. If your history is too thin to support a claim, or everything it finds is already covered, it says so and stops.

Either way you keep corpus.db: a durable, source-referenced map of your recurring work that stays useful long after this run.

Hand the zip to your agent

Download automation-discovery.zip ↓

Skill folders live in different places in Claude Code, Codex, and every other agent, and the fan-out options differ per setup. So don't install it by hand: give your agent the zip and have it do the placement and adaptation. Download the zip, then paste this into your agent:

Show the full prompt

Install the automation-discovery skill from ~/Downloads/automation-discovery.zip (adjust the path if I saved it elsewhere).

1. Unzip it and read SKILL.md and README.md so you know what the skill does.
2. Copy the automation-discovery folder into wherever skills live in THIS setup (for Claude Code that is ~/.claude/skills/), and refresh your skill index.
3. Check what it needs: Python 3.10+, and a fan-out lane for parallel workers. Pick the lane this setup actually has — parallel subagents, or sequential as the fallback — and remember that choice for when we run it.
4. Confirm the install by naming the skill back to me, and list which of my work surfaces you can already read (AI session history first).

Install only. Do not read my history or start any discovery run yet — that happens read-only, and only after I explicitly say go.

Works the same for any agent that can read files and run commands. If yours has no skills folder, it can keep the directory anywhere and load SKILL.md at the start of a run.

Run your first discovery

Discovery is read-only until the final build stage, and there are two human gates: you approve the source list before anything is read, and you choose from the offer sheet before anything is built. Start it with one sentence:

Use automation-discovery. Start with my AI session history; my write root is ~/automation-discovery-run.

Expect it to ask about surfaces, exclusions, and what you already automate before it touches anything. A first run takes about an hour, most of it unattended.