Skip to content
WorksBuddy Docs
AgentsTaroFeatures

Risk Prediction

Continuously scans projects for overdue tasks, stalls, velocity drops, and blocked dependencies, with severity tiers and recommended actions.

Risk Prediction continuously scans four categories of project risk and surfaces each one with a severity tier, the named task or person involved, and a specific recommended action.

How it works

Taro monitors four risk signals continuously rather than at sprint end: overdue tasks, stalled workflows, velocity drops, and blocked dependencies. Each detected risk is scored on two axes, how serious it is and how close to the deadline it sits, then classified into Critical, High, or Monitor. A task is considered stalled when no meaningful activity (status change, comment, attachment, or checklist progress) occurs within a configurable window, defaulting to 48 hours for In Progress tasks and 24 hours for assigned, past-due tasks. Velocity drops are measured against a rolling 3-sprint baseline. Each risk carries an action recommendation generated from the underlying data, with its reasoning shown so leads can override confidently. Risks remain tracked until resolved and escalate automatically if ignored.

Key capabilities

  • Four risk categories scanned continuously and presented in one unified view.
  • Severity tiering (Critical, High, Monitor) based on seriousness and deadline proximity.
  • Velocity-drop thresholds: 15 percent triggers Monitor, 25 percent triggers High, 35 percent with mid-sprint scope additions triggers Critical. Measured at team and per-member level.
  • Blocked-dependency mapping that flags the downstream chain and weights external blockers higher than internal ones.
  • Action recommendations with reasoning for each risk.
  • Automatic escalation of dismissed or unresolved risks, plus a 12-hour follow-up alert for dismissed Critical risks.
  • Configurable thresholds per workspace, including stall window, velocity tiers, blocker escalation age, and scope-creep limits.

A risk cannot be made to silently disappear. You can mark a risk as accepted with a reason, which removes it from the active feed while keeping it in the audit history.