OPEN SOURCE · TOOLS · LABS
Tools you can inspect. Ideas you can try.
I contributed Codex CLI support to Sensei and built early MVPs for two Ketrium Labs tools. Explore the merged work, inspect the tools, or try the hash-chain lab.
Start with the merged Sensei contributionSensei
I added Codex CLI support and changed how Sensei gathers context for code reviews. The merged pull request includes the implementation, tests, and review discussion.
Killcord
I authored Killcord’s initial MVP and later added protection against pending actions being replaced or reversed, best-effort ledger-event correlation, and documentation for a local-only pause server. The tool checks agent actions against configured limits and leaves a tripped action for human review.
Agent Activity Ledger
I authored the initial event-ingestion, timeline, and demo MVP. The tool links each record to the previous one so maintainers can check the chain for breaks; a separately saved checkpoint helps detect a rewritten history. Later contributions added Claude Code transcript import and fixes for idempotency and API-contract edge cases.
From contribution to an interactive model
The lab below turns the public hash-chain idea into a small synthetic exercise: inspect a changed event, then see why a separately saved head matters if a local writer can rewrite the entire chain.
Systems labs
I built these labs to make system behavior easier to examine. Change incoming load, retry behavior, or a cache setting and watch the result. They are repeatable teaching models, not replicas of production systems.