Added a full skill-flow for features/bugs/etc in an existing project

[graviaDaemon]

I've based this skill flow off of the blog by Boris Tane: https://boristane.com/blog/how-i-use-claude-code/
The skill-flow works by letting claude write to $CLAUDE_PROJECT_DIR/.claude/agents/*.md files (adjust the folder path as you prefer in your own projects)

• first by calling /research to allow Claude to read everything deeply, or add a feature argument to let it read only the relevant files, or explain a bug, or that there are bugs, and claude will find all bugs.
• Second you run /plan-it which steps into a planning and annotation cycle to allow claude to plan the implementation, giving you time to read the plan, and write annotations **note** note-description into the markdown file directly. Then when you're satisfied with the plan, you write **APPROVED** at the top of the file.
• Thirdly and lastly, you use /implement, which has claude write a detailed todo list in the plan document, which you can annotate to as well, and finally when you write **APPROVED** at the todo list, claude will implement all changes as planned.

I've been testing this flow in all of my projects, and it is an amazing way of running. If I have constraints for specific tech-stacks, it's quite easy to add more.

For me this is a much stronger means of using claude, since you retain control on what is being changed and how.

[poshan0126]

@graviaDaemon Really appreciate you contributing, these skills are seriously impressive, loved going through them.

The core goal of this repo is to minimize token usage, and I’m noticing Claude Code’s plan mode seems to align pretty closely with that philosophy.

I’m going to take some time to dive deeper and give this PR a proper review soon !