What problem does it solve?
Teams drowning in raw exception volume struggle to tell a genuine regression from background noise. This Skill watches PostHog $exception activity for meaningful changes — bursts, stuck retry loops, multi-fingerprint clusters, and resolved issues that start firing again — and files a validated, deduplicated report only when a finding clears the bar.
Core Features & Use Cases
- Signal-vs-noise discrimination: Uses the
count vs distinct_users relationship on $exception events to separate broad-reach regressions from single-user quirks and retry storms.
- Pattern detection: Identifies fresh bursts, stuck loops, multi-fingerprint clusters with shared root causes, and status regressions on previously resolved issues.
- Report authoring with dedupe: Authors or edits inbox reports via the report channel, using scratchpad memory keys (
pattern:, noise:, dedupe:, report:) so recurring issues are edited rather than duplicated.
- Use Case: A deploy introduces a per-request server error. The scout spots
count ≈ distinct_users spiking on one fingerprint, correlates the stack-trace activity name with a recent deploy via activity logs, and files a P1 report with hourly evidence and a chart.
Quick Start
Ask the Signals agent to run the error tracking scout against your PostHog project and report any fresh $exception regressions worth acting on.