PROC-3 · Fact-checker retention/onboarding

> Item derived from the fact-checking cycle analysis (**2026-07-03**). Production numbers (namespace `main`), unless noted.

### Finding (evidence)
39% of fact-checkers do a single activity and never return; the operation depends heavily on a few (1 account concentrates most initial reviews).

### Action
Improve the first-task funnel (mentoring, simpler entry tasks, fast feedback).

### Impact
Broadens the active base and reduces the concentration bottleneck.

---
- **Priority:** 🟡 Medium
- **Dependencies:** depends on **INST-7** (onboarding milestones) to measure the effect.
- **Epic:** #2490

Start by reading INST-7 and epic #2490 to understand the onboarding milestones and measurement dependency, then review the fact-checking cycle analysis cited in the issue. The work is done when a concrete first-task funnel plan covering mentoring, entry tasks, and feedback is defined with measurable retention and concentration outcomes.

Rédigé par le modèle d'indexation à partir du texte de l'issue.