mlsanityData to 14 Oct 2026

The Compiler May Read It, the Agent May Not: Keeping Part of a Research Code Away from a Coding Agent

arXiv 2609.35557 ↗First posted 28 Sep 2026cs.CR cs.AI cs.SEcited by 0

Not on the map. Why?

Abstract (arXiv)

The compiler must read modules a physics-based solver cannot build without; the coding agent must not read that intellectual property. The harness, the software around the model that executes its tool calls, does not ship that rule. We classified fifteen read routes against a container, permission rules and a sandbox. None of the three can tell which program is reading.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 5 of its 15 references that are in the map. First posted on arXiv 28 Sep 2026.

Builds on

5 of 15 references.

References · 5

  1. 2025Defeating Prompt Injections by Designcited by 297
  2. 2025AgentSpec: Customizable Runtime Enforcement for Safe and Reliable LLM Agentscited by 223
  3. 2025Securing AI Agents with Information-Flow Controlcited by 121
  4. 2025Contextual Agent Security: A Policy for Every Purposecited by 59
  5. 2026Lingering Authority: Revocable Resource-and-Effect Capabilities for Coding Agentscited by 8

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Who is scientific code for? Maintaining human-readable landmarks in agent-written code ↗
  2. 2026When Does Restricting a Coding Agent to executecode Help? A Regime × Agent-Design Ablation
  3. 2025Agent READMEs: An Empirical Study of Context Files for Agentic Coding
  4. 2026What Context Does a Coding Agent Actually Need to Act?
  5. 2026Steerability via constraints: a substrate for scalable oversight of coding agents
  6. 2026Do AI Agents Really Improve Code Readability?
  7. 2026LLM Agent-Assisted Reverse Engineering with Quantitative Readability Metrics
  8. 2026Subjects, Not Authors: The Authorship Hazard in Agentic Dataspaces