Teaching Skills — a teaching-side sibling suite, architecturally inspired by ARS

[YujxZJCN]

Hi Cheng-I — thank you for ARS; its architecture taught me how to build this.

I teach at a university and spent the past period building teaching-skills: a suite for the teaching side of academic life — course design → lessons → assessment → delivery → reflection — built explicitly on the patterns I learned from ARS: SKILL.md + agent ensembles, a shared contract layer, staged gates with human checkpoints, and a passport as the single source of truth (a "Course Passport" instead of the Material Passport). The README credits ARS up top.

The two suites compose at one seam: teaching-reflector's sotl mode (scholarship of teaching and learning) designs a classroom inquiry and then hands off to ARS deep-research / academic-paper for the publication phase.

If you think it would serve your users, I'd be glad to PR a one-line companion link into your README (similar to the experiment-agent slot). Either way — thanks for open-sourcing the architecture. Attribution and license notes are in my README.

[Imbad0202]

Hi, thank you for this. A teaching-side suite built by someone who actually teaches is exactly where I hoped these patterns would travel, and I am glad the ARS architecture was useful as a starting point. I hope ARS keeps serving you well on the research side too; the SoTL handoff into deep-research / academic-paper is a natural seam between the two suites.

A one-line companion link PR is welcome. Go ahead.

One thing from ARS's own path, offered as encouragement rather than critique: the layer that ended up mattering most over time was the machine-checkable one, a JSON Schema for the passport, CI lints, tests around the invariants. If teaching-skills grows that layer around the Course Passport, it will compound. Good luck with the suite, and thanks for the careful attribution.

[YujxZJCN]

Thank you — PR opened: #427, kept to one line in the Companion section as offered.

And thanks for the pointer on the machine-checkable layer. That matches my instinct after assembling the showcase example (cross-file consistency was the hardest part to keep honest by hand), so I'm taking the advice directly: a JSON Schema for the Course Passport, a passport validator with the cross-reference invariants, an executable implementation of the alignment gate's checklist, and CI over a golden fixture are going in next. The lesson that the deterministic layer is what compounds is very believable from ARS's changelog.