Orchestrate multi-wave rewrites with parallel subagents and wave-boundary commits.
npx skills add https://github.com/shawnpetros/claude-skills --skill boil-the-ocean
Please help me install this Agent Skill. Skill: boil-the-ocean Source: https://github.com/shawnpetros/claude-skills/tree/main/boil-the-ocean Command: npx skills add https://github.com/shawnpetros/claude-skills --skill boil-the-ocean
Large rewrites often fail from uncontrolled scope, coordination drift, and context overload, causing broken integrations, missed verification, and unrepeatable outcomes.
In your Claude Code workflow, tell the user/agent to run a multi-day rewrite using the boil-the-ocean skill only after you have a committed plan at notes/PLAN.md, an isolated worktree, and defined build/lint/test gates plus telemetry tracking.
High-intent search queries and answers about installing and using this skill.
Large-scale code rewrites with parallel agents require multi-wave orchestration, worktree isolation, and commit-per-wave discipline to prevent coordination drift. Adversarial review and live-test gates catch integration cracks early, keeping complex changes reviewable and rollbackable.
Multi-day refactors need wave-boundary commits, parallel subagent execution with hard concurrency caps, and per-agent deliverables on disk. This prevents context overload and rate-limit cascades, turning a sprawling rebuild into a repeatable methodology.
Preventing rate-limit cascades in parallel subagents requires hard concurrency caps and worktree isolation. Wave-boundary commits and live-test gates for external systems ensure verification failures are caught and contained within individual execution waves.
Git worktree isolation is enforced for autonomous multi-module rebuilds to keep parallel agent changes separate. Combined with committed planning at notes/PLAN.md and defined build/lint/test gates, it ensures changes remain reviewable and rollbackable.
Parallel agents for large rewrites are not suitable for tasks under two days or lacking a committed plan. You need defined build/lint/test gates, telemetry tracking, and explicitly authorized large-scope autonomous work to justify orchestration over turn-by-turn supervision.
Adversarial review in multi-wave rewrite orchestration catches integration cracks early by evaluating per-agent deliverables at wave boundaries. It pairs with evidence-based retrospective artifacts to turn each run into a replayable, reusable methodology.
JoeCP17Create isolated Git worktrees for feature branches with baseline tests.
telum-aiSynthesize epic retrospectives into a project-retro.md report with cross-epic patterns.
riftzen-bitRun build, type, lint, test, security, and diff validations with a structured report.
whichguyPrompt Hermes models via 'hermes chat -q' for analysis and synthesis.
