This repository collates the work of various Advanced Research + Invention Agency Safeguarded AI Creators.
The Agent Profiles CLI (agentprofiles-cli) enables software developers to easily manage per-project configuration profiles of LLM agent tools via direnv, simplifying the process of working with AI coding assistants across multiple contexts. Published to GitHub and npm.
The Clones CLI (clones-cli) makes it trivial to clone remote Git repositories locally for read-only reference, and keep them up-to-date, without the overhead of managing them manually. Useful for ensuring full context is available to agents when working with third-party libraries, or Git-hosted information repositories (like the Safeguarded AI Commons). Published to GitHub and npm.
Petrinaut is a web-based environment for working with Petri Nets, including those with various extensions.
Previewable at demo.petrinaut.org; published to GitHub and npm.
- 10 Jan 2026 - Reviewing state of the art in Agentic AI
- 17 Feb 2026 - The case for Petri nets for Agentic AI
- 23 Mar 2026 - CatColab v0.5: Sandpiper
Note
Links to the below exploratory design projects will be added shortly.
In progress:
- Agentic Assistance Flows
- Spec Elicitation Flows
- Unstructured Ingestion
- Petri Net Composition
- Petri Net Diffing
Planned:
- Petri Net Abstractions
The Safeguarded AI Commons contains third-party publications and research relevant to the Safeguarded AI programme. This can be cloned into easily-queryable local context via the Clones CLI tool.
The Safeguarded AI outputs above have been developed by HASH, Ink & Switch, the Future of Programming Lab, and others (coming soon).
A full list of the programme's creator teams can be found on the ARIA website.