Claude Code skills: setup and workflow comparison

The speaker walks through six Claude Code skills, plugins and built-in commands for business automation work, plus a frontend design bonus. The overview explains their different jobs and gives installation guidance for the coding assistant.

Skill Creator turns plain-language requirements or an SOP into a reusable skill. Superpowers adds planning, tests before implementation and separate reviews for specification compliance and code quality. The speaker stresses that human QA still matters. GSD assigns tasks to fresh sub-agents to reduce context degradation; he cautions that those agents consume tokens rather than necessarily saving them.

The review section contrasts /review, described as running locally, with /ultra review, which uploads a branch to a cloud sandbox. The stated requirements for the latter are Claude Code 2.1.86 or later and a signed-in Claude account; an API key alone does not work. The speaker says it can take 10 to 20 minutes in the background.

Context Mode filters tool output and records session events in a local SQL database. Claude-Mem stores summaries in SQLite for retrieval across sessions, with a local web viewer. Their efficiency figures come from project benchmarks cited by the speaker. He recommends marketplace installation for Claude-Mem because installing its SDK alone does not register the hooks. Local storage in these workflows does not establish offline model inference.