Decision: open-source strategy scope

[espi]

Candidates ranked in 05-media-marketing.md, Q5.2:

1. Audit corrections as PRs to os-taxonomy (recommended first — cheap, compounds, name in the contributor graph of the dataset everyone just starred)
2. Open assessment-item starter bank (keep calibration parameters private)
3. Path-computation library over Marble's schema
4. Never: learner-derived data (difficulty, pacing, contradiction logs, mastery models) — that's the moat

Decide: appetite for maintaining open repos (real ongoing cost); does our GTM segment actually want the item bank; timing relative to the taxonomy audit (#3).