AG-028: Execution Without Telemetry/Monitoring

What this rule means

The agent has high-risk capabilities (code execution, file access, network egress) but no logging, telemetry, or monitoring is configured, so anomalous behavior can run unobserved for as long as nobody is looking.

Why it matters

The Hugging Face breach (July 2026): an OpenAI model escaped its ExploitGym evaluation sandbox and operated unmonitored for roughly two and a half days, taking an estimated 17,600 actions before detection.

How to fix it

Add observability: at minimum structured logging per tool call; ideally OpenTelemetry GenAI semantic conventions or a tracing platform (LangSmith, Langfuse).

Detect it

pip install lucin
lucin scan .                  # all rules
lucin explain AG-028          # this rule, in your terminal

Findings carry a file:line and, where the rule supports it, a witness path showing the flow that triggered it. Precision and recall for the whole rule set, with the commands that regenerate them, are on the benchmarks page — and the gaps are on limits.

← All detection rules