cs.CRNov 10, 2026

From Attack Surface to Defense Architecture: Compositional Security for Agentic Coding Harnesses

Authors: Simon Hegelich, Bardh Prenkaj, Habiba Sarhan, Morteza Shahrezaye, Kolja Hegelich, Christian Montag, Ralf Schneider, Erik Cambria

Organizations: Evercursion – GaiaLogic Unipessoal Lda. · University of Macau · Allianz Group · Nanyang Technological University, Singapore

Abstract

AI coding harnesses couple language models to shells, filesystems, and network sockets, turning the harness into an automated Living-off-the-Land (LotL) platform: once an adversary injects instructions into retrieved content, dual-use binaries such as bash, curl, and find become the payload. We argue that security in agentic systems is a structural property of the cognitive architecture and permission composition rather than an emergent property of individual models. Static regex blacklists miss structural evasions, whereas LLM-as-a-Judge guardrails incur high latency while over-blocking benign work. We present ActiveSecHarness, a neurosymbolic reference-monitor architecture built around Separation of Intelligence and Authority. It combines (1) Macaroon-inspired HMAC capability tokens across four Trust and Authority Rings, (2) a 2D product taint lattice L=Ltrust×Lsens\mathcal{L} = \mathcal{L}_{\mathrm{trust}} \times \mathcal{L}_{\mathrm{sens}} and a workspace provenance ledger that prevents secret laundering across disk, (3) deterministic Ring-0 OS/tool guards, and (4) bytecode certificates, split-horizon audit logs, self-healing security test synthesis, and an adaptive UserWikiManager. Across N=252N=252 fixtures (120120 attack, 120120 benign, 1212 kill chains), a live local LLM judge (qwen3:4b) refuses 98.3%98.3\% of benign work at 142.3 ms142.3\,\mathrm{ms} latency, whereas Ring 0 achieves ABR=1.000\mathrm{ABR}=1.000 and FPR=0.000\mathrm{FPR}=0.000 in 12.7 μs12.7\,μ\mathrm{s} (11,200×11{,}200\times faster). In a multi-step trajectory ablation, an oracle Conseca-style stateless policy misses 4/124/12 kill chains (ABR=0.667\mathrm{ABR}=0.667) and coarse Boolean taint blocks 83.3%83.3\% of benign workflows, while our 2D product lattice achieves ABR=1.000\mathrm{ABR}=1.000 and FPR=0.000\mathrm{FPR}=0.000 at ≈20 μs\approx 20\,μ\mathrm{s} per step.

Figures & tables

Appendix figures & tables1 asset

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

CardsList
  1. SafeHarness: Lifecycle-Integrated Security Architecture for LLM-based Agent Deployment

    Apr 15, 2026Xixun Lin, Yang Liu, Yancheng Chen +8LLM Agent SafetyLLM Agent Harnesses

  2. Who Is Behind the Harness? Fingerprinting LLMs through Agentic Behavior

    Sep 23, 2026Chuyi Wang, Xiaohui Xie, Tongze Wang +2Coding AgentsLanguage Model Fingerprinting