An evidence-first harness for agent-assisted research
-
Updated
Oct 22, 2026 - TypeScript
An evidence-first harness for agent-assisted research
Spec Kit extension: state-externalizing research harness (budgeted exploration, evidence curation, claim verification) based on Harness-1 (arXiv:2606.02373)
Gated research harness: commit and submit gates, provenance, cluster submission — the scaffolding with the project removed
Self-contained, MIF-native AI research harness template — a goal-driven multi-agent engine, adversarial falsification gate, knowledge graph, and research→report/blog pipelines. Instantiate and update-propagate with Copier.
A self-organising community of LLM agents with no boss: they argue about pizza, review each other by name, and write their own norms.
Track B SONIC SIM curriculum research fork: bounded eval, variant checkpoint provenance, and SIM-M2 gates for GR00T-WholeBodyControl
OpenClaw-only modular autoresearch harness. Runs Karpathy-style question → experiment → evidence loops with deterministic supervision, quality review, safety gates, and reversible implementation handoff.
Provider-independent harness for auditable multi-agent research with competing hypotheses, isolated review, and frozen evidence ledgers.
Portable, evidence-gated research harness and Iterations 1-6 records for the RPCD open problem
To associate your repository with the research-harness topic, visit your repo's landing page and select "manage topics."