link-drop-pipeline

Overview

The meta-pattern: you see a good URL (tweet, repo, blog post, YouTube video) and want it captured into your toolkit RIGHT NOW without a 6-hour R&D session. You drop the URL, the agent does everything: fetch → classify → extract → dedupe against existing 600+ skills → write new skill/agent/command/KB/MEMORY/CLAUDE.md updates → refresh router registry → run autoresearch-evolve on 3-5 related skills → single commit → rsync to Achilles → log decision trail.

This replaces the monthly R&D marathon (captured in docs/snapshots/rd-session-2a672502-handoff.md) with a 5-minute per-URL primitive. When aggregated across a week of daily drops, it recreates the marathon outcomes incrementally.

When to use

• You just saw a tweet/repo/post worth capturing
• You don't want to stop what you're doing to build a skill
• You want the toolkit to grow passively, not from sprint-style marathons
• You have Gemini key + X bearer token + Achilles SSH working
• router-hub is built (reads registry)
• autoresearch-evolve is available (to propagate learnings)

Workflow