Selected work
Public projects, inspectable evidence.
Two portfolio projects connecting technically complex systems with the information people need to evaluate them. Explore their capabilities, source material and boundaries below.
Gleipnir
Agentic AI Guardrails Plugin
Coding-agent work needs a reviewable record of its instructions, repository changes and validation. Gleipnir provides local evidence collection and advisory checks around that workflow.
- Who it is for
- Developers working with coding agents
- What you can inspect
- The public repository describes an append-only evidence ledger for task authority, commands, agent claims and approvals. Its default workflow is passive, and its README documents local installation and validation commands.
- Status and limits
- The README labels it a local-only developer preview. Its heuristics do not prove correctness or replace tests, security review or human review; it is not a permission or deployment-verification system.
Celebtir.com
Onchain DeFi Market Intelligence
Protocol risk depends on who can change code, pause operations, alter parameters or influence token supply. Celebtir presents evidence-linked assessments to help readers examine those control arrangements.
- Who it is for
- Readers examining DeFi control and trust assumptions
- What you can inspect
- The public project provides protocol comparisons and control graphs, with methodology, data contracts and historical snapshots available in its repository. Its live interface offers a way to inspect the assessments.
- Status and limits
- The assessments are not a smart-contract security audit, a combined safety score or an investment recommendation. Coverage differs across systems, and historical repository snapshots are not a live feed.