architecture
01Command Claude Code
Evaluate an architecture proposal — deep-context take (full system) vs first-principles (proposal only) — to catch sunk-cost rationalisation before the change lands.
Cross-session debate protocol — same model, different context depths, better decisions
Command Claude Code
Evaluate an architecture proposal — deep-context take (full system) vs first-principles (proposal only) — to catch sunk-cost rationalisation before the change lands.
Command Claude Code
Pressure-test a deprecation decision — builder's view (knows why it exists, who depends on it) vs cold reader (just sees the thing and the question "should this go?") — to catch attachment-driven maintenance.
Command Claude Code
Hiring decision second opinion — manager's view (full résumé + interview context + team fit) vs identity-blind qualifications-only view — to catch familiarity / pedigree bias in candidate evaluation.
Command Claude Code
Run a context-asymmetric debate on the current question — Deep (full project context) vs Fresh (zero context) — and return the convergence result.
Command Claude Code
Run a Ploidy debate on a post-mortem — context-rich account (knows history, people, prior judgements) vs timeline-only reader — to surface blame anchoring and unexplored root causes.
Command Claude Code
Evaluate a product decision — maker's view (knows users, history, constraints) vs first-time-reader view (spec only) — to catch sunk-cost feature bias and "because we always do it this way" product choices.
Command Claude Code
Run a Ploidy debate on a pull request — deep-context reviewer vs zero-context reviewer that sees only the diff + description — to catch over-engineering, scope creep, and rationalised author bias before merge.
Command Claude Code
Sanity-check a design idea BEFORE implementation — runs a Ploidy debate between the full-context view and a zero-context fresh sub-agent to surface rationalisation before code gets written.