OMX-style convergence interview with ambiguity scoring, readiness gates, and explicit handoff artifacts. Use when: requirements are vague, scope is unclear, multiple interpretations exist, the user says "I'm not sure what I need", or before starting any large feature. Produces a canonical spec/state pair that ralplan, ralph, and team consume as binding upstream context.
Persistent completion loop with formal phase state machine. Execute → verify → fix cycle that keeps pushing until the task passes or max iterations reached. Use when: implementing features that need verification, fixing failing tests, completing multi-step work with quality gates, or any task that benefits from an automated verify/fix cycle.
Multi-reviewer consensus planning workflow. Draft a plan, then iterate through architect and critic reviews until both approve. Use when: planning complex features, reviewing architecture tradeoffs, creating implementation plans that need validation, or any work that benefits from structured plan approval before execution.
Read-only reconnaissance mode. Explores the codebase without modifying anything. Use when: understanding a new codebase, finding where something is implemented, tracing a code path, discovering patterns, or answering "where is X?" questions.
Language-aware bounded shell commands. Run safe inspection and verification commands without full agent autonomy. Use when: checking syntax, running linters, viewing routes, inspecting git state, or any quick shell command that should be bounded and safe.