Agent skill · marketing growth · jeremylongshore
posthog-incident-runbook
Triage a production PostHog integration incident while preserving application availability and evidence. Use when capture, flags, private API access, or downstream destinations are degraded. Trigger with "PostHog incident", "PostHog outage", or "PostHog on-call".
Why this skill is useful
Adds executable scripts for incident triage and error handling specific to PostHog that aren't in public documentation.
What it needs
Requires bash installed locally. Requires posthog account access. About 5k tokens when loaded. Last updated 2026-08-07. 2,604 stars on the source repository.
What this skill does
PostHog Incident Runbook Overview Rapid incident response for PostHog integration failures. PostHog Cloud has its own status page (status.posthog.com) — the first step is always determining whether the issue is PostHog-side or your integration. Prerequisites The affected service, deployment window, and PostHog region are known. Read-only evidence is preferred; production write probes require explicit authorization. A safe application fallback exists for analytics and flag failures. Instructions Tool discipline Use Read to inspect the relevant configuration and implementation before proposing changes. Use Grep to locate initialization, capture, flag, and credential boundaries. Follow the triage sequence below. Stop when evidence identifies a failed boundary; do not continue mutating unrelated layers. Severity Levels Level Definition Response Time Examples ------- ------------ --------------- ---------- P1 Analytics completely down < 15 min All capture calls failing, feature flags returning defaults P2 Degraded analytics < 1 hour High latency, partial event loss, slow flag eval P3 Minor impact < 4 hours Webhook delays, specific event type missing P4 No user impact Next day Monitoring gaps, dashboard stale data Quick Triage (Run First) Do not use an event capture as the default health check: it writes project data, and an HTTP 200 only confirms receipt and payload shape, not successful ingestion. If the incident commander explicitly authorizes a production write probe, use a named synthetic event and distinct ID, record the approval and timestamp, inspect quotalimited, and remove or exclude the probe from analysis. …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills jeremylongshore/posthog-incident-runbook