Hey PH — Ivan, solo founder. Started from a personal pain: I'd close a B2B deal or set up a process for a consulting client, and the context died with the session. Next client — start from zero. It's 3 Claude skills: • /openexp-pack-author — records the raw trace of a session that worked • /openexp-pack-pii-check — anonymizes it (only needed if you share outside) • /openexp-use — matches a pack to your current situation and replays it First version I shipped had baked-in conclusions ("lesson 1, lesson 2..."). Then someone asked "why these conclusions and not others?" Fair. Rewrote it to ship the raw trajectory instead. Each reader's agent derives its own lessons. Or uses mine. It's basically SOPs / playbooks for AI agents. 5 real packs published, all mine: • 3 B2B sales acquisitions (57-72 days each, day-by-day) • 1 OSS publishing run • 1 GA setup on a static site These packs are raw — first iteration, my own runs only. The format will only get better as more people publish and push back on what's missing. I want feedback as much as anyone — what's broken, what's confusing, what's worth keeping. Want to bounce ideas, get feedback on a pack idea, or just say hi — book a slot: calendly.com/ivanpasichnyk/30min MIT, runs locally, Claude-native. Catalog: openexp.ai · Repo: github.com/anthroos/openexp What did your AI help you ship this month?

OpenExp

Save, reuse, share AI agent runs that worked

Save, reuse, share AI agent runs that worked