productionos-auto-swarm-nth

Iterative code review automation uses nth-iteration agent swarms to systematically cover a codebase, spawning parallel waves, evaluating outcomes, and re-swarming gaps until achieving 10/10 quality. It applies across software review and code-quality tasks by chaining skills and enforcing guardrails.

Recursive agent orchestration coordinates parallel agent waves to evaluate and repair code, measuring outcomes per wave before re-swarming remaining gaps. This mechanism ensures complete codebase coverage and target quality through repeated cycles of evaluation and targeted fixes.

Automated code repair workflows support configurable guardrails including max_waves, swarm_size, and max_cost limits. These constraints ensure safe, deterministic execution by bounding the scope and expense of each recursive agent swarm run.

ProductionOS supports optional worktree isolation for agent swarms, ensuring safe and repeatable execution during codebase evaluation and repair. This prevents conflicts when parallel waves apply fixes within the same repository.

Re-swarming is ideal for large codebases requiring iterative review and validation across multiple quality dimensions. A single agent pass often misses gaps, whereas nth-iteration swarms evaluate outcomes per wave and re-swarm uncovered areas until complete coverage.

Launch the nth-swarm workflow by invoking the Codex-native alias productionos-auto-swarm-nth on your repository. This initiates the recursive swarm orchestration to systematically evaluate, repair, and validate your codebase.