Skip to content
View GvHildebrand's full-sized avatar
🎯
Focusing
🎯
Focusing

Block or report GvHildebrand

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
GvHildebrand/README.md

Gregorio von Hildebrand

I set up AI agents inside businesses and teach the people there to run them. I live in Costa Rica.

Every agent I build is a folder of plain text files. You can open it, read what each step does, and change it without calling me. The method is the Interpretable Context Methodology by Van Clief and McDermott.

I run my own land business, Dear Wise Earth, on the same method.

How I run agents

I run all my ventures with one AI chief of staff and a team of project agents, built in plain folders. These are the public versions of that setup.

Repo What it does
agent-chief-of-staff A folder template for one Claude Code assistant that runs every project you have. It reads an inbox first, then checks that every scheduled job actually ran, then picks up where the last session stopped. Its memory is plain markdown, and it keeps a log of its own mistakes.
claude-code-skills Seven skills that make the agent check before the expensive mistakes: a pre-mortem on any plan, a five-line declaration before anything runs unattended, memory upkeep, sales copy and UX review.
open-work-log Turns your Claude Code sessions into a public daily work log. Anything private is held by default.

Open tools

Repo What it does
sentinel-hook Checks every shell command and file write a Claude Code agent attempts, then allows, asks or blocks it. Each decision goes into a tamper-evident log. No AI model sits in the blocking path.
negotiator Reviews a contract clause by clause. Prices each clause as money at risk, and prices every edit by what it costs the other side too.
eu-ai-act-article-50-auditor Audits an AI product against Article 50 of the EU AI Act. Every finding quotes the line of law it rests on.
cr-parcel-record Turns a Costa Rican parcel's registry report, survey plan, water concession and deed into one record where every value cites the line it came from.
pre-duediligence-skill A first-pass due diligence report on a piece of land, with every claim sourced and tagged by confidence.

I build these with Claude Code. sentinel-hook and the Article 50 auditor are works of Vigilia, a Swiss association in formation that keeps a watch on AI.

What I worked on each day is at sovran.works/log.

Work with me

AI in your business. A private 1:1 session on your real project, and you leave with something working on your own machine. Or a Fractional Chief AI Officer retainer. At sovran.works.

Land in Costa Rica. Start with a free reading of the parcel. Then a call with me for $500, with a written Land Position in five days. At gregorio.dearwise.earth.

Pinned Loading

  1. poor-phill-refund-agent poor-phill-refund-agent Public

    An autonomous refund agent for Claude Code. Reads the thread, replies in your voice, escalates on a schedule, never harasses.

    TypeScript