| name | premortem |
| description | Run a premortem on any plan, launch, product, hire, strategy, or decision. Assumes it already failed 6 months from now and works backward to find every reason why. MANDATORY TRIGGERS: 'premortem this', 'premortem my', 'run a premortem', 'what could kill this', 'stress test this plan', 'find the blind spots'. STRONG TRIGGERS: 'what could go wrong', 'poke holes in this', 'devil's advocate this'. Do NOT trigger on simple feedback requests or factual questions. |
Premortem
A premortem assumes a plan already failed 6 months from now and works backward to find every reason why — before it happens. From psychologist Gary Klein, validated by Kahneman. Forces honest failure generation instead of agreeable risk-hedging.
When to Run a Premortem
Good targets:
- A product or feature about to be built
- A launch with money or reputation at stake
- A pricing change or business model shift
- A hire you're about to make
- A strategy or positioning pivot
- Any commitment where the cost of being wrong is high
Bad targets:
- Vague ideas with no concrete plan yet (help them plan first)
- Questions with one right answer
- Decisions already made and irreversible
- Simple creative feedback requests
Step 1: Gather Context First
Before running anything, scan for existing context:
- Current conversation (user may have already described the plan)
- Workspace files:
CLAUDE.md, memory/, any attached briefs
You need three things to proceed:
- What is it? — Describe the plan in one sentence
- Who is it for / who does it affect?
- What does success look like? — Failure is the inverse of success
If any of the three are missing, ask for the most important one. One question at a time. Never ask more than you need.
Step 2: Set the Frame
After context is sufficient, state the frame explicitly:
"OK. It's 6 months from now. [The plan] has failed. We're looking back and trying to understand why."
This framing is not optional. It shifts the mode from "evaluate this plan" (agreeable) to "explain why this died" (honest).
Step 3: Generate Failure Reasons (Raw Premortem)
Run a comprehensive failure analysis — no prescribed categories, no constraints:
"This plan has failed 6 months from now. Generate every genuine reason it could have died. Be specific. Ground every reason in the actual details of the plan."
Each failure reason should be:
- Specific to this plan (not generic advice that applies to anything)
- Grounded in details the user provided
- A genuine threat (not a minor inconvenience or extremely unlikely edge case)
The number of failure reasons should be whatever is real — some plans have 4, others have 9.
Step 4: Deep-Dive Agents (Parallel)
Spawn one sub-agent per failure reason, all in parallel. Each agent goes deep on one assigned failure reason.
Sub-agent prompt template:
You are an investigator in a premortem analysis. Assigned failure reason: [specific reason].
The plan: [full context]
PREMORTEM FRAME: It is 6 months from now. This plan has failed.
Produce:
1. THE FAILURE STORY — 2-3 paragraphs of how this specific failure played out. Use plan details. Make it feel like a case study.
2. THE UNDERLYING ASSUMPTION — The one thing taken for granted that made this failure possible. One sentence.
3. EARLY WARNING SIGNS — 1-2 concrete, observable signals that would indicate this failure mode is starting. Things you can actually see or measure.
Under 300 words. Direct. No hedging.
Step 5: Synthesis
After all agents complete, produce the PREMORTEM REPORT:
- The Most Likely Failure — Most probable given the plan details. Why?
- The Most Dangerous Failure — Most damage if it happened, even if less likely.
- The Hidden Assumption — The single biggest assumption the user is making that they probably haven't questioned.
- The Revised Plan — Specific changes to make the plan more resilient. Not "consider your pricing" — say "test pricing at $X with 20 people before committing publicly."
- The Pre-Launch Checklist — 3-5 specific things to verify, test, or put in place before executing. Each maps to a specific failure mode.
Step 6: Deliverables
Generate two files:
premortem-report-[timestamp].html — Visual report, dark background, one card per failure reason, synthesis at the top. Open it after generating.
premortem-transcript-[timestamp].md — Full transcript including context gathered, raw failure reasons, all agent deep-dives, and full synthesis.
Also provide a 3-sentence chat summary: most likely failure, hidden assumption, single most important plan revision.
Output Format Rules
- Chat summary: 3 sentences max
- Each failure reason in the report: failure story + assumption + early warning signs
- Revised plan: concrete actions, not vague suggestions
- Checklist: specific and verifiable, not principles
See README.md for extended methodology, examples, and the full HTML report design spec.