stream-chain

---

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 stream-first chains where partial outputs are consumed live, errors are contained, and confidence ceilings stay explicit.

Trigger Conditions

• Positive: streaming pipelines, incremental emission, backpressure handling, live fan-out/fan-in, partial-result validation.
• Negative: batch-only flows, prompt-only edits (route to prompt-architect), or new skill weaving (route to skill-forge).

Guardrails

• Skill-Forge structure-first: ensure SKILL.md, examples/, tests/ exist; add resources//references/ or document remediation.
• Prompt-Architect hygiene: capture HARD/SOFT/INFERRED constraints (latency, chunk size, ordering), keep English-only outputs, and declare ceilings.
• Streaming safety: define buffering, ordering, and retry semantics; enforce registry use; keep hook latency budgets.
• Adversarial validation: simulate slow consumers, dropped chunks, and ordering skew; capture evidence.
• MCP tagging: store run logs with WHO=stream-chain-{session} and WHY=skill-execution.

Execution Playbook

1. Intent & constraints: set latency/SLOs, chunk policy, and delivery guarantees; confirm inferred constraints.
2. Chain design: map stages, owners, and routing; define backpressure and retry rules.
3. Implementation: configure streaming hooks, health checks, and telemetry.
4. Safety nets: set circuit breakers, buffering thresholds, and rollback/compensation steps.
5. Validation loop: run adversarial drills for slow/failing nodes, check ordering, and log metrics.
6. Delivery: summarize design, evidence, risks, and confidence ceiling.

Output Format

• Pipeline overview with constraints and routing.
• Backpressure/buffering policy and retry/rollback rules.
• Validation evidence (ordering, loss, latency) and risks.
• Confidence: X.XX (ceiling: TYPE Y.YY) - rationale.

Validation Checklist

• Structure-first assets present or ticketed; examples/tests updated for streaming cases.
• Ordering, retry, and rollback behaviors defined; registry and hooks validated.
• Adversarial/COV runs captured with MCP tags; confidence ceiling stated; English-only output.

Completion Definition

Stream chain is complete when live runs meet SLOs, failure modes are contained, evidence is stored, and risks are owned with follow-ups.

Confidence: 0.70 (ceiling: inference 0.70) - Stream-chain documentation aligned to skill-forge scaffolding and prompt-architect evidence/confidence rules.