Root Cause Pattern — Context
Pattern: Root Cause
Component: context.md
Version: 1.1 | Updated: 2026-07-16
Preconditions (must be true before activation)
- Incident is mitigated or stable under monitoring (harm contained)
- Incident ID, severity, duration, and impact summary exist
- Raw timeline sources available (alerts, deploys, flags, chat, tickets)
- Facilitator will enforce blameless norms aloud
- Likely action owners attend (or delegate with authority)
- Draft will use the RCA structure from the SDLC prompt (summary, timeline, 5 Whys, detection gap, remediation, prevention)
Trigger conditions (apply the pattern)
Apply when any of the following hold:
- SEV1 or SEV2 mitigated
- SEV3 with clear recurrence risk or customer commitment to a writeup
- Hotfix/rollback fixed symptoms but systemic gap remains
- Near-miss that would have been SEV1 with minor timing differences
Do not apply when
- Incident still uncontained — finish hotfix-pattern first
- Purpose is to punish an individual — wrong process
- No evidence timeline can be constructed (still write what is known; mark gaps; do not invent)
- Purely speculative tabletop with no real event (use game-day notes instead)
Required inputs
| Input | Source | Minimum quality |
|---|---|---|
| Incident summary | Incident ticket | ID, SEV, impact, duration |
| Timeline sources | APM, CD, flags, chat | Timestamps |
| Contributing context | Deploys, traffic, config | Facts labeled |
| Logs/metrics | Observability | Key excerpts linked |
| Mitigation steps | IC timeline | What already ran |
Authority boundary
The RCA agent drafts. Humans in the RCA meeting own the root-cause statement and action assignments. Legal/comms may restrict external versions. The agent must not invent owners or due dates.
Version: AIES v1.0.0✏️ Edit this page on GitHub