multi-agent-orchestrationlisted

# Multi-Agent Orchestration ## Goal Turn a raw backlog + agent roster into an executable multi-sprint plan that respects dependencies, prevents file-ownership conflicts, and enforces the merge order from `docs/runtime/multi-agent-merge-sequence.md`. ## When to use - Planning Phase 5 §5b execution (Waves pattern) with ≥4 specialist agents dispatched - Rescuing a stalled sprint where conflicts emerged from parallel edits - Onboarding a new modernization program — translate `execution-roadmap.md` items into Wave groupings ## Inputs - `backlog` — JSON or YAML array of tasks. Each task: - `id` (e.g., W2.1, NF-03) - `title` + `description` - `effort_estimate` (session-units or hours) - `dependencies` (list of prior task ids) - `owner_agent` (specialist that should claim it, e.g., "backend-api-architect") - `target_files` (absolute paths the task will write) - `sprint_length_days` (default 14) - `num_agents_per_sprint` (default 4, clamped to MAX_PARALLEL_AGENTS = 6) - `agent_capacity_hours_per_day` (default 6; varies by agent — e.g., backend 8, QA 6) ## Outputs - `reports/current/sprint-plan.md` — per-sprint task assignments - `reports/current/dependency-graph.md` — topological view with critical path highlighted - `reports/current/file-ownership-matrix.md` — which agent owns which file in each Wave - `reports/current/merge-sequence.md` — depth-ordered merge plan (infra → backend → frontend → tests) - `reports/current/daily-standup-template.md` — what each agent report