What problem does it solve? LLMs default to polite, hedged risk lists when asked to evaluate a plan. This Skill reframes the question as a past-tense failure narrative ("the plan has already failed — explain how"), producing concrete, specific failure causes instead of generic caveats, and records decisions so they can be revisited later. ## Core Features & Use Cases - Four-phase session: pre-flight context sync, silent scan with three parallel helpers covering six angles, live decision dialog per hole, and atomic persistence to docs/premortem/<plan>.md with history snapshots. - Deterministic mechanics via CLI scripts: plan-type classification, helper-response JSON schema validation, hash-based dedup, stable H-NNN hole IDs, and atomic file writes with locking. - Bias check and conditional reverse-premortem: a 6-item bias checklist (Kahneman/Lovallo/Sibony) on the top picks, plus a reverse-premortem when the recommendation is delay, reduce stake, or abort. - Use Case: Before launching a paid workshop, run a premortem to surface 5–8 concrete failure holes (pricing, audience, competition), pick mitigations for each, and re-run a month later to compare which holes closed or worsened. ## Quick Start Ask the assistant to run a premortem on your concrete launch or decision plan, for example: "Run a premortem on my plan to launch a paid workshop for marketing teams."
