Agent Atlas / Runtimes and workbenches

andrewyng/openworker

A local-first desktop agent runtime whose provider switching, approval gates, reusable skills, and security-focused coworkers make the harness itself an inspectable security boundary.

Desktop agent runtime Scope profile Project documented Reviewed August 26, 2026
Intended operator

People delegating file, terminal, connected-app, and security work from a desktop agent.

Primary surfaces

desktop workbench / local Python agent server / security coworkers and reusable skills

Tracked repository

andrewyng/openworker

Teaching mechanism

Security coworkers drive scanners instead of impersonating them

The bundled code, dependency, and cloud-posture workflows name concrete scanners, require coverage disclosures when a tool is missing, and reserve fixes or cloud changes for reviewable code paths.

Evidence limitOpen instructions make the intended workflow auditable, but do not establish scanner availability, model judgment quality, safe configuration, or production effectiveness on a particular machine.
Inspect the source
System map status

Reporting is still in progress.

This scope profile names the project's place in the field and one useful mechanism. It does not claim that every runtime boundary has been inspected.

Learn Through This Project

9 lessons

Compare Its Boundaries

8 labs

Latest Reporting

0 articles

No Daily Edition reporting is attached yet.

Primary Sources

Open Questions

  • Can an auditable harness keep scanner coverage, model judgment, tool installation, and cloud authority visibly separate for non-specialist teams?

Evidence state describes what this Atlas profile can support today. It is not a product score and it is not evidence that uninspected capabilities are absent.