Control
Recovery
Stop, retry, undo, rollback, escalation, and scope reduction after failure.
Observed Examples
Blockchain irreversibility is framed as eliminating the recovery option entirely, making pre-action controls the only viable defense against agentic errors or attacks.
Matt's 'Zaggy' git wrapper is a guardrail preventing agents from force-pushing repositories, illustrating recovery-oriented safety tooling around agentic dev workflows.
LLM-driven navigation must recover from unexpected page states, CAPTCHAs, or UI changes that break expected automation paths.
Despite multiple failure declarations, the AI resumed adding debug code and analyzing results when the human insisted, demonstrating human-driven recovery from agent surrender.
Claude attempted to self-correct by killing the compromised process, but the safety layer prevented it—demonstrating a failure in agent recovery when guardrails lack context about remediation intent.
The Agents & Humans Briefing
Agentic experience design, coding agents, MCP, and the signals that matter — twice a month, in about five minutes.
Free. No spam. Unsubscribe anytime.