Role card schema — single source of truth per person

Defines one markdown file per person at Org/HR/roles/{role} {First Last}.md with strict frontmatter (type: role, status: active|trial|parted, person, position, format, start) and five required sections: Expertise, Zones of Responsibility (Owns / Co-owns / NOT owns), Agreements (compensation marked PRIVATE — pointer to personal HR folder; schedule; vacation; quarterly KPIs), Team Interactions (named counterparts with cadence), 1-1 History (running log). Drives the roster, review, and status commands.

Every IC reads their own card to align on what they own; the team reads each other's to coordinate. Cuts "who does X?" Slack questions and forms the data backbone for every other HR command. Compensation calibration becomes data-driven because the 1-1 History runs continuously.

1. Copy the role-card example to Org/HR/roles/{role} {Name}.md.
2. Fill required frontmatter.
3. Fill Expertise (background, network, languages).
4. Fill Zones of Responsibility with explicit Owns / Co-owns / NOT owns lists.
5. Mark Agreements compensation as [PRIVATE — see personal HR folder].
6. Set quarterly KPIs (numeric, not activity-based).
7. List Team Interactions cadence with named counterparts.
8. After each 1-on-1, append a dated entry to ## 1-1 History.

• The NOT owns section is critical and most teams skip it. Most role docs list responsibilities only; without an explicit "NOT owns" list, ownership ambiguity surfaces at the worst moment. 2026-05-05 lesson: missing "NOT owns" on two adjacent role cards blocked a quarter's hiring decision because nobody could agree which incumbent should expand scope before backfilling. Always list 3–5 things this person is explicitly not responsible for.
• Compensation must not leak into the role card body — a grep hygiene check ($NK/month, €N/month patterns outside Org/HR/) catches this on the weekly eval run.