Back to the changelog

v0.85.8

August 26, 2026

**Sensitive data stays sensitive and the numbers tell the truth.** An indicator that would reveal one person's data now stays hidden for everyone, impossible durations no longer distort time and cost — and you can see how many were excluded — and processes that wait for several tasks to finish now publish without workarounds.

  • Security

    Indicators on sensitive data protected from every role

    Indicators computed on sensitive fields no longer show the value when that value effectively is one person's data: minimum and maximum metrics, single-case indicators and averages over fewer than 5 cases stay hidden for **everyone**, not just read-only users. Averages over a sufficient number of cases remain visible, and goals on these indicators are still fully manageable. This protects against direct reading, not against deliberate analysis by someone who already has access to the process: where a stronger guarantee is needed, avoid using the sensitive field as the indicator's source, or restrict roles.

  • Improvement

    Time and cost without impossible durations

    A single out-of-scale duration could make a process look months long. Durations beyond the plausibility threshold no longer enter time and cost calculations, and next to the numbers you now see **how many were excluded**, with per-process detail. Historical values may drop as a result: they are moving toward the true figure. The trend chart is built on previously recorded snapshots and realigns over time.

  • Feature

    Processes waiting on several tasks now publish

    An automatic phase that only moves forward once several tasks — even from different phases — are complete is no longer rejected at publish time as having no way out: instead you get a non-blocking suggestion to add a safety timeout. And a case legitimately waiting on those tasks no longer triggers the stalled-automation alert.

  • Security

    Tighter control over outbound connections

    The cross-installation collaboration channel now verifies the destination of every outbound call, not just some, and validates the verification-keys address before storing it. Internal-address detection was hardened in three cases where an internal network address could slip through: the protection applies to every outbound channel, from webhook integrations to AI providers.

  • Design

    A hidden value says it is hidden

    Where an indicator is protected, a placeholder with an explanation takes the place of the number, instead of meaningless percentages or empty charts. A protected history is no longer shown as "insufficient history", and a goal with a hidden outcome no longer appears as "Missed".