Ask Claude to Kill Your Plan Before You Fall in Love With It
A pre-mortem prompt surfaces the failure you're too excited to see. I run it on every plan now.
Before I commit to a project, I paste the plan into Claude and ask it to write the post-mortem for the version that failed. Not 'what could go wrong' in a soft way. I ask it to assume the project is six weeks dead and explain exactly why. The answer is usually the thing I was already avoiding.
The prompt matters. 'Find the three most likely ways this fails and be specific' beats 'give feedback.' Specificity forces it off generic caution into the actual weak joints of my plan: the dependency I hadn't confirmed, the person I assumed would help, the metric I couldn't actually measure.
I then argue back. The point isn't to kill good ideas; it's to find the one fatal assumption. Twice it saved me from launching something that would have embarrassed me in a month. Once it confirmed the plan was sound, which is its own kind of useful.
Run this before the kickoff, not after the first setback. A pre-mortem is cheap; a mid-project pivot is not. Make it a habit and your plans get quietly better without you noticing why.