Monokit is on sale: the app screens most kits leave out · See Monokit → · Read the post →

← Studio

Engineering · 2026

The Research Problem I Solved with a Slash Command

A Claude Code skill that turns URL pasting into three persistent artefacts: a public essay, an internal signal log, and roadmap backlog items.

Update, November 2026: I no longer run this skill. The post stays as a record of the idea.

I read about LangGraph on a Tuesday. Relevant to the agent work, so I made a mental note and closed the tab. By Thursday it was gone: not forgotten entirely, just decayed into a vague sense that there was something worth looking at.

That kind of loss doesn't feel expensive, which is the problem. Interesting threads erode quietly while I'm building something else.

So I built a Claude Code skill to catch them: /research-capture.

What it does

I drop one or more URLs into the command. The skill fetches each page and turns it into three things:

  • A public essay in /blog. Not a summary of the links: what the technology is, why it matters structurally, and what it changes about how I'd build.
  • An entry in docs/RESEARCH-SIGNALS.md, a private, permanent log. Stack implications for my own agent setup (Moirai, MakerSuite, Claude Code), with concrete integration points.
  • Rows in docs/ROADMAP.md under Future, each linking back to the essay.

Why it lives in the IDE

The IDE is where I already have context. I'm mid-session, I find something, and my instinct is to note it and get back to work. The skill fits that instinct instead of fighting it.

It deliberately doesn't decide whether to act. Not everything worth capturing is worth building now, so the skill keeps the option open without forcing the call.

It's global (in ~/.claude/skills/, not one repo), so any session in any project can run it. Over time the signal log becomes a map of where the work is heading, the roadmap items link back to the evidence, and the essays become a public record of how my thinking moved.

It pays off over months, not sessions.