I have an embarrassingly large number of articles in my Readwise Reader queue.
I installed the CLI documented: MCP & CLI Tools - Readwise Docs - there is an MCP if you prefer that, they have the same functionality. (I prefer the CLI’s because they use fewer tokens most of the time.)
Then I installed the skills that they offer as a starting point: GitHub - readwiseio/readwise-skills: Agent skills for your Readwise and Reader data, powered by the Readwise MCP server/CLI. Triage your inbox, quiz yourself on what you've read, build a personalized now-reading page, and more. · GitHub and ran the build_persona skill.
It built a decent persona (a snippet from it):
Taste & Sensibility
- Skeptical-pragmatist on AI. Will save and highlight Mollick and Narayanan and Choudary. Wants both the optimistic operator view and the critical research view, then reconciles them. Has no patience for cheerleading.
- Evidence-first. Strong pull toward research and data: arXiv (13 saves), Nature, HBR, SpringerLink, Frontiers, Scientific American. Will read a primary-source paper before he’ll read a hot take about the paper.
- Practitioner voices over pundits. John Cutler, Ben Linders, Klaus Leopold, Teresa Torres, Christoph Steinlehner, Jason Yip, Matt Parker, Bryan Finster. Tends to trust people who have shipped product or coached teams.
- Pragmatic on marketing. Reads sales/marketing books for craft, not for vibes. Pulls out specifics (“hire a storytelling coach to elevate that pitch”, “kill lower-performing ads monthly”).
- Curious about cognitive science, not pop psychology. Pentland, Waber, Rock, Oakley, Buzsáki — empirical or at least research-adjacent. Less interest in self-help abstractions.
- Comfortable with long-form. Books and PDFs are normal; not chasing short-form quick wins.
…Then I asked it run a triage on my Later queue, skipping articles that were already referenced in my vault.
Only 650 articles to go.