# Pre Mortem

> Imagines the plan has already failed six months in the future, then works backwards to enumerate every reason why — surfacing risks that optimism and momentum normally hide. Use this skill when the user is finalizing a plan, launch, migration, hire, or big decision and asks "what could go wrong?", "review my plan", "what am I missing?", "poke holes before we commit", or is about to commit to something expensive or hard to reverse. Pairs with grill-me (which stress-tests at the start); pre-mortem assumes failure has happened and reasons backward.

- Skill: `opelpleple/pre-mortem` (Agent Skill)
- Install (CLI): `npx skillmds@latest add opelpleple/pre-mortem`
- Raw SKILL.md: https://api.skillmd.com/api/skills/opelpleple/pre-mortem/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: opelpleple (https://skillmd.com/u/opelpleple)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/opelpleple/pre-mortem

---


# Pre-Mortem

Assume it's six months from now and the project failed. The question isn't "could it fail?" — it's "it failed; now, why?" That framing unlocks risks people won't admit while they're still hopeful.

## When to use
- Finalizing a plan, launch, migration, or roadmap before committing.
- High-stakes, hard-to-reverse decisions (hire, vendor lock-in, public launch).
- The team is confident and aligned — exactly when blind spots hide.
- Right before a go/no-go gate.

## When NOT to use
- Early ideation where you're still generating, not committing.
- Trivial or easily reversible decisions (don't pre-mortem a font choice).
- The team is already demoralized — this can tip into doom; reframe toward mitigation fast.
- You have no plan yet — there's nothing concrete to fail.

## The method
1. Set the scene out loud: "It's [date, ~6 months out]. The project shipped but failed / never shipped. It's a clear, undeniable failure." Make it vivid and concrete.
2. Generate failure stories independently first — list as many distinct ways it died as you can before discussing any. Quantity beats polish here.
3. Cover the categories: technical, people/team, scope/timeline, external/market, dependencies, assumptions that turned out false, and "the boring thing nobody owned."
4. For each failure, write the one-line obituary: "It failed because ___."
5. Score each on likelihood x impact. Focus on high-impact, plausible failures, not exotic ones.
6. For the top risks, name the earliest warning sign — what would you observe in week 2 or 4 that this is happening?
7. Convert each top risk into a mitigation or a tripwire: prevent it, or define the signal that triggers a course-correction.
8. Output: ranked risks, leading indicators, and concrete pre-commitments (what you'll do now, what you'll watch for).

## What good looks like
- At least one risk surfaces that nobody had said aloud before.
- Risks are specific ("the third-party API rate limit kills our import at scale"), not generic ("scope creep").
- Each top risk has an early warning sign and an owner.
- The plan changes as a result — a mitigation added, scope trimmed, or a tripwire set.

## Anti-patterns
- Vague risks ("communication issues") that can't be watched for or prevented.
- Listing only risks you already planned to handle (theater, not discovery).
- Stopping at the scary list with no mitigations or tripwires.
- Pre-mortem as veto — using it to kill a project rather than de-risk it.
- Only technical risks; the killers are often people, ownership, and false assumptions.

## Example
Plan: "Launch the new billing system on the 1st."
Pre-mortem (it's the 1st of next quarter; the launch failed):
- "It failed because customers were double-charged during the cutover." (High impact, medium likelihood.) Early sign: mismatched totals in the dry-run reconciliation. Tripwire: any reconciliation delta > $0 in staging blocks launch.
- "It failed because the old system's edge-case discounts weren't migrated and finance found out in week 3." Early sign: discount count differs between systems. Mitigation: diff discount records before cutover; assign an owner.
- "It failed because nobody owned customer comms and support drowned." Sign: no comms draft exists 1 week out. Mitigation: assign comms owner today.
Result: launch gated on zero reconciliation delta, a discount diff added to the checklist, and a named comms owner.