Plugins

Compares coding agents in isolated Git worktrees with deterministic validation, scoring, and explicit winner application.

Run the same prompt through every model you have, and see the difference without squinting.

Evidence-first multi-model coding comparison workbench for DeepSeek Harness

Evidence-backed Skill self-evolution for DeepSeek Harness: redacted trajectory learning, quarantined proposals, arena validation, canary monitoring, automatic rollback.