The automation trap: I kept spending 2 hours to build 5-minute tasks. So I fixed it.

You know exactly what you want the workflow to do. Five steps. Twenty minutes of actual logic. But you spend two hours on: Setting up API auth Wiring error handling Figuring out cron syntax Writing the boilerplate to read env vars Making the scheduler actually reliable The workflow logic takes 20 minutes. The surrounding scaffolding takes 2 hours. I had 10 of these workflows sitting on my "automate eventually" list for months. So I built skillforge — tell it what you want in plain English, get production-ready Python + a deployment config back. How a session looks $ skillforge new "every morning at 9am, summarize my open GitHub PRs and post to Slack" Compiling skill... Trigger: cron ( 0 9 * * 1-5 ) Steps: 1. fetch_prs — GitHub API, list open PRs 2. summarize — Claude, generate standup summary 3. post — Slack, post to #standup Generated: ~/.skillforge/skills/morning-standup/ skill.yaml skill.py Run it now? [ y/N]: What you get Two files, both human-readable: skill.yaml — the manifest: n