Risks What could happen Mitigation
Identity drift Same person saved as multiple stakeholders under different names Use stakeholder IDs as canonical; prefer link ... to STK-...; periodically merge duplicates in 👥 Stakeholders
Wrong links A task/context row gets linked to the wrong stakeholder Keep explicit link commands; review Stakeholder IDs during weekly maintenance
Incomplete graph Some tasks/context rows stay unlinked, so reasoning is patchy Build a habit: when assigning or logging important context, add stakeholder links at the same time
Over-trust in relational reasoning The system may sound more relational than it truly is Treat current stakeholder-aware reasoning as assistive, not authoritative; keep key human judgment in loop
High-priority execution risk CoS could act outward on something important too early Keep the explicit confirmation gate for high-priority CoS tasks via confirm T###
Workspace sync mismatch Google/Smartsheet mirrors may not have matching columns or may miss optional fields Keep schemas aligned; use dedicated sync sheets if your operational sheets are complex
Privacy leakage Stakeholder notes may accumulate sensitive details in shared sheets Keep notes operational and minimal; avoid storing private or unnecessary personal information
Cache staleness Cached prompt results may survive slightly longer than ideal TTLs are short and input-based; if needed, rerun after meaningful changes or reduce TTLs
Rejected-signal misclassification “Why not now” reason may be plausible but imperfect Use the batched reason classifier plus heuristics; review 🪵 Rejected Signals periodically
Schema migration friction Older sheets with manual edits can drift from expected columns Auto-migration covers the common case; still worth checking tabs after upgrade