I have been doing two things with Claude Code lately.
Step 1: Research the thing that bothers me
I saw a competitor selling well. First reaction: anxiety. But before giving anxiety the wheel, I wanted to figure out what the thing actually is.
I used a Skill called “H/V analysis”:
- Vertical — how did it walk step by step to its current shape?
- Horizontal — compared to its peers, where are the differences?
Once these two dimensions are filled in, anxiety gets replaced with information. I now know whether the competitor is “actually strong” or “good at marketing”, and which gaps I should care about versus which ones I can ignore.
Step 2: Get clear about myself
After seeing the outside clearly, the next step is looking back at myself.
This is where I use the first-principles Skill:
- The thing I am doing — what is the bottommost “why”?
- Which constraints are real, and which are just “everyone else does it this way”?
A lot of the time, the source of anxiety is not that the competitor is good. It is that I have not thought my own thing through clearly.
How the two Skills relate
- One helps you see the outside
- The other helps you think the inside
Used separately, both work. Used together, they complete each other.
Both are public on GitHub, and any Claude user can install them:
- H/V analysis: https://github.com/KKKKhazix/khazix-skills
- First principles: https://github.com/danyuchn/first-principles-skill
A workflow detail
I do not use these Skills inside a “real work” session. I use them in a research-only session before plan mode, then collect the conclusions into notes. When it is time to actually do the work, I open a new session and bring the polished conclusions in.
Separating “research” from “execution” prevents the long-context attention dilution problem.