ThumbGate 1.28.0 Adds Safer Rule Enforcement and Interactive Setup for AI Agents

ThumbGate 1.28.0, an open-source AI-agent safety tool, addresses the challenge of converting human feedback into enforceable rules without disrupting legitimate workflows. The update introduces an interactive quickstart command that guides operators through installing their first prevention rule, while five new slash commands make it easier to inspect active rules, blocked actions, and system readiness. A key design choice separates negative feedback from positive: explicit 'never' corrections can trigger immediate enforcement, whereas 'always' preferences remain advisory since they rarely define a safe denial boundary on their own. Before any candidate rule is promoted to enforcement, the system replays it against previously approved actions, quarantining rules that would have blocked known-good behavior. The release also adds a Hermes Agent adapter, strengthens package-integrity verification, and hardens protections against environment overrides and process-kill attempts.