1:1 & review prep skill

One CLI command reads a person's role card (with ## 1-1 History), their personal HR folder, and recent tasks from the task tracker — then generates a pre-filled review agenda. Supports four review types: Monthly 1-on-1, Quarterly Review (with compensation), Annual Review (with 360 feedback), Fit Assessment (end of Month 1 for new hires). For unusually large raises the founder can layer on a second pass: same data, two prompts — "why this person deserves the biggest raise" vs "why no raise, possibly part ways" — both with citations.

Removes the "what should we talk about?" friction at every 1-on-1 and calibrates compensation discussions against actual quarterly data. The raise-evaluation overlay catches false positives at 30+ headcount where the founder doesn't directly work with the person daily.

1. Maintain role card with disciplined ## 1-1 History entries — date, wins, blockers, next-month focus.
2. Run /team-hr review Anna.
3. Skill reads role card, personal HR folder, task tracker (Linear MCP).
4. Skill outputs agenda: Wins (5m) / Blockers (10m) / KPI progress (10m) / Next month scope (10m) / Open questions (5m), pre-populated with data points.
5. For Quarterly: add KPI scorecard, compensation review, growth path (title, equity, new directions).
6. For raise calibration: run two opposing prompts, both demanding citations from the person's outputs (Slack/calls/tasks/commits).
7. After the meeting, log a dated entry to ## 1-1 History.

• Compensation discussion artifacts must live in the personal HR folder Org/HR/{Person}/, never in the role card body — the role card is read by anyone running /team-hr roster and comp leaks have a cultural blast radius. Treat the raise-evaluation output as input, not authority: spot-check cited evidence (LLMs will hallucinate "commits" that don't exist).