cascade-orchestrator

---

LIBRARY-FIRST PROTOCOL (MANDATORY)

Before writing ANY code, you MUST check:

Step 1: Library Catalog

• Location: .claude/library/catalog.json
• If match >70%: REUSE or ADAPT

Step 2: Patterns Guide

• Location: .claude/docs/inventories/LIBRARY-PATTERNS-GUIDE.md
• If pattern exists: FOLLOW documented approach

Step 3: Existing Projects

• Location: D:\Projects\*
• If found: EXTRACT and adapt

Decision Matrix

Match	Action
Library >90%	REUSE directly
Library 70-90%	ADAPT minimally
Pattern exists	FOLLOW pattern
In project	EXTRACT
No match	BUILD (add to library after)

---

STANDARD OPERATING PROCEDURE

Purpose

Design and operate cascaded workflows where each stage consumes prior outputs, enforces quality gates, and exposes controlled escape hatches for recovery.

Trigger Conditions

• Positive: multi-stage pipelines, gated reviews, dependency-driven workflows, staged rollouts, evidence-capturing handoffs.
• Negative: single-step tasks, linear prompts without gating (route to prompt-architect), or new skill creation (route to skill-forge).

Guardrails

• Skill-Forge structure-first: maintain SKILL.md, examples/, tests/; add resources/ and references/ or log follow-ups.
• Prompt-Architect hygiene: document intent, stage-by-stage HARD/SOFT/INFERRED constraints, and provide pure-English outputs with ceiling-aware confidence.
• Stage safety: define entry/exit criteria per stage, cache artifacts, and include rollback/skip rules that prevent cascading failure.
• Adversarial validation: simulate partial failures, retries, and out-of-order delivery; capture evidence and timing.
• MCP tagging: persist cascade runs with WHO=cascade-orchestrator-{session} and WHY=skill-execution for traceability.

Execution Playbook

1. Intent & constraints: collect objective, success metrics, and required sequencing; confirm inferred dependencies.
2. Stage design: enumerate stages, owners, inputs/outputs, and gate criteria with timing budgets.
3. Routing plan: wire registry-approved agents, define escalation paths, and set retry/backoff rules per stage.
4. Safety nets: pre-mortem failure points, define rollback/skip logic, and guard shared state.
5. Validation loop: run dry-runs and adversarial drills; log evidence, timings, and deltas.
6. Delivery: present cascade map, gate outcomes, residual risks, and confidence ceiling.

Output Format

• Cascade overview with stages, owners, and gate criteria.
• Constraint matrix (HARD/SOFT/INFERRED) and risk register.
• Routing rules (retries, rollbacks, escalation paths).
• Validation evidence with timings and artifacts; remaining risks/open items.
• Confidence: X.XX (ceiling: TYPE Y.YY) - rationale.

Validation Checklist

• Structure-first assets ready or ticketed; examples/tests updated or queued.
• Gate criteria defined and met per stage; retries/backoffs verified.
• Registry-only agents used; hooks within latency budgets; rollback tested.
• Adversarial/COV results stored with MCP tags; confidence ceiling declared.

Completion Definition

Workflow is complete when each stage passes gates, artifacts are persisted, recovery paths are validated, and cascade telemetry is recorded with MCP tags.

Confidence: 0.70 (ceiling: inference 0.70) - Document aligns cascade orchestration with skill-forge structure and prompt-architect evidence and confidence rules.