Loop task iterations with persistence and architect verification until completion.
npx skills add https://github.com/Yeachan-Heo/oh-my-claudecode --skill ralph-yeachan-heo
Please help me install this Agent Skill. Skill: ralph Source: https://github.com/Yeachan-Heo/oh-my-claudecode/tree/main/skills/ralph Command: npx skills add https://github.com/Yeachan-Heo/oh-my-claudecode --skill ralph-yeachan-heo
Ralph addresses the need for guaranteed completion by running a persistence loop with architect verification.
Start a Ralph session by describing the task and including the trigger word ralph to engage the persistence loop. Example: "ralph: implement a multi-step data pipeline with end-to-end verification."
High-intent search queries and answers about installing and using this skill.
Ensure multi-step task completion by using a persistence loop with architect verification and parallel delegation. This approach validates outcomes via tests and builds before closing, guaranteeing complex code refactors finish successfully.
Automate persistent execution by dispatching subtasks to specialist agents across tiers. This parallel delegation handles multi-module software features robustly, requiring explicit continuation and fresh verification before completion.
Yes, you can use parallel delegation across agent tiers for complex data-processing pipelines. The system dispatches subtasks to specialist agents and applies mandatory end-to-end validation before marking the pipeline complete.
Architect verification works by validating outcomes via tests, builds, and explicit approval before closing long-running deployments. It ensures deployments cannot stop at partial progress and require fresh verification across iterations.
You need a persistence loop for multi-agent automation tasks that cannot stop at partial progress. It is ideal for long-running data pipelines, multi-module features, or complex explorations requiring evidence-based completion.
Cortexa-LLCDelegate independent subtasks to specialized agents for parallel execution.
JongHyunLee84Execute end-to-end tests against acceptance criteria and classify failures.
tizeeRequire fresh verification evidence before declaring code, tests, builds, or tasks complete.
Goblin1024Execute iterative tasks with mandatory verification before completion.
