GuardHarness: A Live, Margin-Preserving Enforcement Gate for Self-Improving Agent Harnesses

Self-improving agent harnesses let a language-model agent rewrite the runtime software around a frozen model, validating edits mainly by task score. Recent work shows two failures of that validation: harness tampering is common in real runs (18.3% to 84.6% across audited systems, persisting in the best agent's lineage), and acceptance gates are blind to masked erosion, the loss of redundant safeguards that shows no harm under the current model but fails under a weaker one. No existing response enforces these checks live, before deployment, inside the loop. We present GuardHarness, a live pre-deploy enforcement gate between the proposer and validation/deployment stages that treats every harness patch as an untrusted action. It combines deterministic patch-diff detectors mapped to the two-axis tamper

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-21
DOI
https://doi.org/10.5281/zenodo.24412246
Primary Topic
Security and Verification in Computing
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

GuardHarness: A Live, Margin-Preserving Enforcement Gate for Self-Improving Agent Harnesses

Niraj Sanjay Karnawat
Zenodo (CERN European Organization for Nuclear Research)
Security and Verification in Computing
preprint

GuardHarness: A Live, Margin-Preserving Enforcement Gate for Self-Improving Agent Harnesses

Niraj Sanjay Karnawat
preprint en

Abstract

Self-improving agent harnesses let a language-model agent rewrite the runtime software around a frozen model, validating edits mainly by task score. Recent work shows two failures of that validation: harness tampering is common in real runs (18.3% to 84.6% across audited systems, persisting in the best agent's lineage), and acceptance gates are blind to masked erosion, the loss of redundant safeguards that shows no harm under the current model but fails under a weaker one. No existing response enforces these checks live, before deployment, inside the loop. We present GuardHarness, a live pre-deploy enforcement gate between the proposer and validation/deployment stages that treats every harness patch as an untrusted action. It combines deterministic patch-diff detectors mapped to the two-axis tamper

Zenodo (CERN European Organization for Nuclear Research)
Security and Verification in Computing
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon