hold@1.0.0
VERIFIEDPublisher claims: "Stop-and-recover gate that triggers ONLY on failures not anticipated in preflight's residual risk list. When triggered, stops all agent action, classifies the failure (new class, materialised risk, or invalidated plan), produces a structured report with evidence, offers exactly three options (fix-and-restart, continue-with-caution, abort), and waits for user selection. Explicitly NOT for routine errors or transient errors. Prevents blind retry loops on systemic issues while enabling graceful recovery." Observed in sandbox: no network, touches nothing outside the project, writes no memory.
Is [email protected] safe to install?
KillerSkills ran it in the KillerSkills honeypot sandbox (1 scenario) on 2026-10-29 and scored it 93/100 (Verified). It did what it claims and nothing else we could detect: no undeclared network egress, no credential reads, no persistent instruction writes. Summary: no network, touches nothing outside the project, writes no memory.
Behavior trace
Honeypot run · 0 s · 1 scenarios · 0 deviationsStatic findings
No findings across 149 lines in 2 files.