Pre-compaction memory checkpoint. Use right before compacting a long session (context heavy, ~150k+) to curate durable knowledge into the project's long-term memory BEFORE the conversation is summarized away. Forces fact extraction, dedup against existing memory, protocol-correct writes, and an index update — then hands off to /compact. Does NOT run /compact itself (that's a user command).
Systematic bug-hunting mode with logic reasoning. Use when user wants to find bugs before they happen — combines pattern-based detection (fast) with logic reasoning (deep) to find structural defects, design flaws, and edge cases. Forces 2-phase scan, severity rating, and actionable report before any fix.
Pre-deploy simulation mode. Use when a fix is written but not yet verified — to step-trace the code through normal/edge/failure scenarios, expose what was missed, and require explicit case coverage before commit. Pairs with rootfix.
New-feature build mode that respects existing patterns and never merges without tests. Use when user wants to add a new feature (not fix bugs, not refactor) — new API endpoint, new module, new functionality. Forces requirement clarification, existing-pattern survey, design proposal, and test-coupled implementation before any code is written.
Refactor-only mode that preserves behavior while removing code smells. Use when user wants to clean up code (not fix bugs, not add features) — duplication, long functions, naming inconsistency, magic numbers. Forces baseline capture, pattern recognition, behavior-preserving transforms, and commit separation before any change.
Root-cause-first debugging mode. Use when user wants a real fix (not a band-aid) — when a "fix" keeps coming back, when same code path is built in N places, or when async paths interact in non-obvious ways. Forces 5-Why drilling, drift/race detection, and trade-off-explicit options before any code change.