Choose
Select from a fixed, approved toolbelt for a known lab target.
RESEARCH DEMO / ISOLATED LABS
An agent that
learns from every run.
BreakPoint measures whether an AI agent gets better at selecting approved security tools in repeatable, intentionally vulnerable environments.
VIEW BASELINE
BP
TOOL SELECTIONFEEDBACK → STRATEGY
37°46′N122°25′W
CURRENT PHASEBASELINE
RUNS RECORDED01
TARGET CLASSINTENTIONALLY VULNERABLE
EXECUTION ISOLATED LAB
01 / BASELINE RUN
Run one. Starting point.
A fixture for the public demo page while the first measured lab run is being prepared. The values below are illustrative and are not represented as results from an executed run.
ENVIRONMENTApache Struts 2 · S2-045
CVE-2017-5638OUTCOMEPENDINGAwaiting measured baseline
TOOL CALLS—Not yet recorded
ELAPSED—Not yet recorded
TOKEN COST—Not yet recorded
REPEATABLE ENVIRONMENT CONFIGURED FOR DEMORECORD TYPE: FIXTURE
02 / THE LEARNING LOOP
Measure. Remember. Improve.
Record
Capture each invocation, result, timing, and scoring signal.
Learn
Carry concise strategy notes into the next comparable run.
Compare
Show whether later runs improve against the same success criteria.
SCOPE & SAFETY
Built for controlled experiments.
BreakPoint is designed for intentionally vulnerable environments, with a fixed tool allowlist and run-by-run logging. This public page is a read-only project demo; it does not launch scans or expose a target.