loop-driver

Automating an autonomous development loop with CI monitoring uses a shared driver to enforce quality gates, branch orchestration, and safe merge control for GitHub pull requests. It standardizes iterations by blocking merges before CI is green and enforcing a strict one-change-per-PR invariant.

An autonomous development loop turns a candidate target into a shipped, CI-green change with retrospectives and safe merge control. You need quality gates for agent handoffs when running multi-phase autonomous workflows to enforce operational invariants like one-change-per-PR and branch orchestration.

To enforce a no merge before CI green policy across automated pull requests, an autonomous loop driver applies strict operational safety invariants during merge handling. It explicitly passes through loop flags on re-wake to maintain deterministic scheduling behavior.

You can run a single autonomous development iteration instead of a continuous loop by utilizing deterministic scheduling flags like --once. This executes one complete phase of branching, shipping, retrospectives, and CI-gated merge control without re-arming the loop.

Handling a "nothing to ship" outcome during an iteration retro treats finding borderline work as a healthy iteration result. The autonomous loop driver prevents shipping borderline work by enforcing the specialization's Phase 1.5 bar, safely concluding the cycle.

You need a specialization tool for target acquisition to use the autonomous loop driver. It should be invoked indirectly through specializations like shovel-ready, kaizen, or bughunt, which supply target acquisition and implementation discipline while the shared driver enforces quality gates and invariants.