Agent skill · product strategy · vishalsachdev
premortem
Run a premortem on a plan, launch, product, hire, strategy, or decision — assumes it failed 6 months from now and works backward to find every reason why, then produces a revised plan. Use when the user has a concrete plan or commitment with high cost-of-being-wrong and asks to "premortem", "stress test", "find blind spots", "poke holes", "what could kill this", or "what am I missing". Skip for vague ideas without a plan, simple feedback requests, factual questions, or already-irreversible decisions.
Why this skill is useful
Provides a structured framework for conducting premortems, enabling the AI to generate specific failure analyses that it wouldn't typically produce on its own.
What it needs
About 4k tokens when loaded. Last updated 2026-08-01. 4 stars on the source repository.
What this skill does
Premortem Imagine the plan already failed. Work backward to find why. Based on Gary Klein's HBR method — the "this is dead, explain how it died" frame produces more specific, honest failure modes than "what could go wrong?". This breaks Claude's default agreeableness: instead of looking for reasons the plan works, you're explaining how it fell apart. --- When to run Good targets: product/feature launches, pricing changes, hires, strategy pivots, partnerships, any commitment where being wrong is expensive. Bad targets: vague ideas (help plan first), factual questions (just answer), draft feedback (that's editing), already-irreversible decisions. If the user wants multiple perspectives on a current decision rather than failure analysis, suggest a different tool — premortems specifically simulate post-failure hindsight. --- Step 1 — Gather minimum context You need three things before running: 1. What is it? — describable in one sentence 2. Who is it for / who does it affect? 3. What does success look like? — failure is success inverted Scan first, ask second. Check the conversation, then run at most 2-3 Glob/Read calls for: CLAUDE.md, memory/ folders, files the user referenced, related project briefs. If something's missing, ask one focused question at a time. Don't interrogate — infer when you can. --- Step 2 — Set the frame explicitly Once context is sufficient, say plainly: "It's 6 months from now. [The plan] has failed. We're looking back to understand what went wrong." This framing is the mechanism. Don't skip it. --- Step 3 — Generate failure reasons Produce a comprehensive list of genuine failure reasons. Each: Specific to this plan (not generic) Grounded in details the user provided A real threat (not a minor edge case) State each in 1-2 sentences. The number is whatever's real — could be 3, could be 9. Don't pad. Don't truncate. …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills vishalsachdev/premortem