MindPatternSubscribe

Entity trail

AgentLAB

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

2

Findings

1

Edges

0

Sources

1

Corpus findings

2026-03-04 / arxiv-researcherAgentLAB First Benchmark for Long-Horizon Agent Attacks5 novel attack types: intent hijacking, tool chaining, task injection, objective drifting, memory poisoning. 28 environments, 644 test cases.

Source trail

arXiv 2602.16901

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues