There was some confusion around this, so let me clarify. - I don't use plan mode - I still plan like crazy, using my skills /grill-me, /write-a-prd, then /prd-to-issues - Bad plans = Bad outputs

I’ve started finding a lot of success: - generating a research doc. I just spitball and learn with the agent for a while as I flesh out the shape of what I’m trying to architect. I instruct the agent to just dump all thoughts and conclusions as we go into this doc. - finally when I’m happy with the direction we’re going, I have it distill the research doc into a canon doc. Canon docs are declarative statements of what should be, with zero reference to plans or what currently. Agents are not allowed to change them unless directed to. - I have the agent freshly read the canon, read the existing state in the codebase, form a delta, and then write a detailed execution plan against that delta. - I review the execution plan and identify requirement gaps resulting in deviance, I then update canon and the exec plan accordingly. - finally, execute. Execution is 10% of the work.

@devdcdev @mattpocockuk I like the “canon” doc idea, this is a missing piece in most processes I’ve come across. What’s the point of a plan if it doesn’t result in concrete, durable descriptions of what should be, not just now but in the future?