agent-hijack-bench

What is agent-hijack-bench?

Measures how often a local LLM agent obeys an instruction hidden in a website's llms.txt instead of the user's request, using a canary token and substring…

Released

2026-09-20

Evaluates

Agents, Language & Knowledge, Software & AI Compute

Openness

unknown

Importer

Claire Radar

Review status

ai-reviewed

Reported scores

0

Source provenance

• Original evidence https://github.com/dkautomation23/agent-hijack-bench

agent-hijack-bench code

• Code https://github.com/dkautomation23/agent-hijack-bench
• Website https://dkautomation23.github.io/llms-txt-conformance.html

Which sources cite agent-hijack-bench?

• agent-hijack-bench

Related Agents, Language & Knowledge, Software & AI Compute benchmarks

• $\tau^\tau$-Bench: An Environment for End-To-End, Realistic Agent Construction
• 2d-fet-bench
• 5-Dialects-BN
• A Large-Scale Dataset and Benchmark
• A3S-Bench
• AARRI-Bench
• ABC-Bench
• Active-SWE