Work in progress. This catalog is under active development and has not been finalized.
Adrian's personal skill catalog for Claude Code — process and knowledge skills reused across projects.
Written for and tested with Claude Code only. The installer can target other agents (Cursor, Codex, …), but nothing here is adapted or verified for them.
npx skills add Adrian333Dev/flow-skillsThe picker lets you choose which skills to install, which agents to install into, and the scope (-g/--global for your user; default is the current project).
Grab one skill directly:
npx skills add Adrian333Dev/flow-skills --skill <name>Skills are symlinked by default, so git pull keeps them current. Each skill carries a CHANGELOG.md — after pulling, check it to see what changed. Use --copy for a frozen copy instead.
brainstorm— map a decision space before acting: build a branch tree, walk it one branch at a time, record every decision in full; sub-files carry the flow through spec and implementation plan.capture-context— write important information to the repo the moment it surfaces: decisions, findings, future ideas, session checkpoints. Passive habit plus on-request routing.debug-web-pages— investigate and reverse-engineer live web pages you don't control: capture snapshots, probe runtime state, run live experiments. For browser-extension, scraper, and userscript work.execute— execute a plan.md task by task: mechanical tasks delegated to cheap subagents by default, verification after every task, inline mode for direct execution.explain— how to explain a proposal, design, or mechanism to someone seeing it for the first time: whole picture first, every component defined from zero, priority-ordered depth, structural ASCII diagrams only.research— never work against an external tool from stale memory: fetch current docs via llms.txt with local caching, escalate to source-code investigation, delegate heavy reading to subagents, external-LLM prompts for landscape research.
Authoring conventions live in CLAUDE.md.