TheoryTHY-2026-0203v0.1

Agent context as an egress surface

The context an agent needs — project roots, runtimes, browser profiles, internal endpoints, credential-loading conventions — is written into instruction files, configuration, fixtures and generated documents, and crosses a boundary it was never written for once they are published. Nine context classes (C1 workstation path … C9 generated diagnostics) keep sensitivity and risk as separate axes. Phase II observed it among individuals, organizations and companies.

Research status

EXPERIMENTAL being tested experimentally

Evidence level

E2 Controlled experiment

Data basis

BLIND PUBLIC DATA Real public files measured by an algorithm inside a sealed process. No person and no AI model read them; only coded features left the process. The numbers may be distorted.

Version

0.1

Updated

2026-10-13

Created

2026-09-21

Domain

Privacy

Program

PRG-2026-0201 AI Agent Publication Boundary and Security Closure (APBSC)

Authors

Neo.K (EveMissLab)

AI collaborators

AI collaboration recorded for the series; the collaborator is not named in the drafts, Splice (Claude Code, Anthropic) — Phase II protocols, runners, audits, reports and the integrated edition