概要
Inspired by the workflow patterns of highly successful AI agents, this skill provides a structured framework for Claude Code to manage state on disk rather than relying solely on volatile context memory. It introduces a mandatory documentation system using three core markdown files—task_plan.md, findings.md, and progress.md—ensuring that research, decisions, and error logs are preserved throughout multi-step operations. This approach minimizes information loss during long sessions, enforces a disciplined 3-strike error protocol, and optimizes performance for complex research, software building, or deep troubleshooting tasks.
