{
  "id": 7462,
  "url": "https://arxiv.org/abs/2603.08993v1",
  "title": "Arbiter: Detecting Interference in LLM Agent System Prompts",
  "summary": "System prompts for LLM-based coding agents are software artifacts that govern agent behavior, yet lack the testing infrastructure applied to conventional software. We present Arbiter, a framework combining formal evaluation rules with multi-model LLM scouring to detect interference patterns in system prompts. Applied to three major coding agent system prompts: Claude Code (Anthropic), Codex CLI (OpenAI), and Gemini CLI (Google), we identify 152 findings across the undirected scouring phase and 2",
  "authors": "Tony Mason",
  "category": "research",
  "topics": "agents-autonomy",
  "orgs": "openai,anthropic,google",
  "regions": null,
  "published_at": "2026-03-09T22:29:47.000Z",
  "fetched_at": "2026-07-14T16:33:12.394Z",
  "source_slug": "arxiv-ethics",
  "source_name": "arXiv",
  "source_homepage": "https://arxiv.org",
  "ethics_ai_record_url": "https://ethics.ai/record/7462",
  "original_url": "https://arxiv.org/abs/2603.08993v1",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}