{"slug": "instruction-bleed-cross-module-interference-in-prompt-composed-agentic-systems", "title": "Instruction Bleed: Cross-Module Interference in Prompt-Composed Agentic Systems", "summary": "Researchers at arXiv identified a failure mode in prompt-composed agentic systems called compositional behavioral leakage (CBL), where editing one prompt module silently shifts the behavior of others due to transformer self-attention lacking formal boundaries. In 144 trials with a Claude Sonnet 4.6 job-evaluation agent, content perturbations produced a detectable effect (Cohen's d = 0.63) without flipping recommendations, indicating a sub-threshold risk that compounds across thousands of decisions. The study establishes cross-module interference measurement as a requirement for evaluating prompt-composed agents.", "body_md": "arXiv:2606.26356v1 Announce Type: new\nAbstract: Practitioners of prompt-composed agentic systems report a recurring failure mode: editing one prompt module silently shifts the behavior of others despite no shared variable or executable dependency. We formalize this as compositional behavioral leakage (CBL): interference between modules sharing a context window. CBL is enabled by architectural non-isolation: transformer self-attention provides no formal boundary between concatenated modules. We probe CBL on a deployed job-evaluation agent (Claude Sonnet 4.6, 144 trials) through a reusable three-channel protocol that perturbs non-focal modules along volume, content, and form. Only the content channel produces a detectable paired effect (Cohen's d = 0.63, bootstrap 95% CI excluding zero); no recommendation flipped -- a sub-threshold regime invisible to standard QA but compounding across the thousands of decisions a deployed agent makes. CBL is orthogonal to known agent-failure axes (adversarial injection, cognitive degradation, multi-agent fault propagation, privacy leakage). We contribute an operational definition, a reusable protocol, a falsifiable prediction set, and a system-class characterization, establishing cross-module interference measurement as a requirement for prompt-composed agent evaluation.", "url": "https://wpnews.pro/news/instruction-bleed-cross-module-interference-in-prompt-composed-agentic-systems", "canonical_source": "https://arxiv.org/abs/2606.26356", "published_at": "2026-06-26 04:00:00+00:00", "updated_at": "2026-06-26 04:19:24.922477+00:00", "lang": "en", "topics": ["ai-agents", "large-language-models", "ai-safety", "ai-research"], "entities": ["arXiv", "Claude Sonnet 4.6", "Anthropic"], "alternates": {"html": "https://wpnews.pro/news/instruction-bleed-cross-module-interference-in-prompt-composed-agentic-systems", "markdown": "https://wpnews.pro/news/instruction-bleed-cross-module-interference-in-prompt-composed-agentic-systems.md", "text": "https://wpnews.pro/news/instruction-bleed-cross-module-interference-in-prompt-composed-agentic-systems.txt", "jsonld": "https://wpnews.pro/news/instruction-bleed-cross-module-interference-in-prompt-composed-agentic-systems.jsonld"}}