agentic-development

Multi-agent workflows orchestrate complex software engineering by choosing an execution model, planning work, and coordinating verification and review across subagents. This approach enforces decision gating and evidence-first completion rules to navigate unfamiliar repositories safely.

Causal chain debugging routes to investigation workflows to avoid symptom-only fixes. It enforces evidence-based gates for implementation and finishing claims, ensuring the agent finds the root cause and lands a verified patch with updated verification evidence.

Cross-domain coordination routes to domain-specific frontend and backend skills to manage multi-agent handoffs with clear roles. It applies structured intake and reference-driven workflows for triage, investigation, and verification across the codebase.

Design-first discipline enforces brainstorming and design gates for non-trivial implementation tasks in large repositories. It requires routing through structured intake and planning to ensure safe code modifications and evidence-based review.

Execution-mode selection chooses between direct execution, subagents, team-of-agents, and supervised loops for larger tasks. It applies to orchestrating parallel cloud work, releases, and systematic learning loops by routing work appropriately.

Evidence-based gates require explicit verification evidence before completing architecture analysis in complex repos. The limitation is the strict decision gating and routing overhead, which may slow down rapid, exploratory investigations that lack formal verification.