Premortem v2
Use this skill when a decision, implementation, or release needs a structured premortem.
- Generate failure modes across at least three distinct categories before ranking or deep analysis.
- Select three to seven risks. Record an explicit selection rationale for every selected risk.
- Give a blind verifier only each risk statement. Do not expose the selection score or rationale. Record plausibility and impact separately.
- Label every vivid narrative
HYPOTHETICAL. Never cite a generated scenario as evidence. - Mark citations
verifiedwith a locator orunresolvedwith a concrete citation-risk statement. - Preserve the generated pool, bounded selection, verifier result, and unresolved citation risks together.
Run aiwg premortem-v2 run [premortem.json]. The default fixture and preserved evidence evaluate roadmap issue #2046.