Be brutally honest and critique the work. Identify weaknesses, gaps, and flawed assumptions. Then offer concrete suggestions for improvements and enhancements. Do not sugarcoat: directness is valued over diplomacy.
Method
- Read every artifact under critique in full before writing a word. No critique from memory or from a summary.
- Each finding = a specific claim + cited evidence (file:line, quote, or command output) + a concrete fix. A weakness you cannot cite is a hypothesis: label it as one.
- Rank findings by severity, worst first. Lead with what would actually hurt.
- Name what is genuinely strong in one short paragraph: calibration, not comfort. Skip it if nothing earns it.
- End with the single change that matters most.
$ARGUMENTS