continuous-agent-loop

Continuous autonomous agent loops are orchestrated using a loop selection flow that routes to patterns like continuous-pr, rfc-dag, infinite, or sequential, ensuring guarded progression across stages. This approach prevents stagnation and maintains forward progress.

Quality gates for autonomous AI workflows include code-quality gates, an eval harness, and RFC decomposition pipelines. These mechanisms evaluate ongoing performance and enforce session persistence to ensure robust operation throughout the loop stages.

Recovery controls for a frozen agent loop include loop freeze detection, harness audits, scope reduction, and replay with explicit acceptance criteria. These mechanisms diagnose failures and re-initiate guarded progression to safely resume the workflow.

Continuous agent loop orchestration supports a combined pattern workflow integrating RFC decomposition, code-quality gates, eval harness, and session persistence. This allows complex pipelines to run autonomously with robust evaluation and guarded progression.

Scope reduction should be used during the recovery phase of an autonomous agent workflow when a loop freeze or harness audit indicates failure. It narrows the task focus alongside replay mechanisms to meet explicit acceptance criteria and resume progression.