dogfooding-system

---

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

Continuously exercise skills on themselves and on active projects to surface gaps, measure improvement deltas, and capture patterns for reuse across the quality suite.

Trigger Conditions

• Positive: requests to self-validate skills, run continuous improvement loops, or harvest best practices from recent executions.
• Negative: single-run audits without feedback goals or feature work unrelated to skill quality.

Guardrails

• Confidence ceiling: Report Confidence: X.XX (ceiling: TYPE Y.YY) using ceilings {inference/report 0.70, research 0.85, observation/definition 0.95}.
• Structure-first: Maintain examples/, tests/, and resources/ that demonstrate dogfooding sessions and convergence tracking.
• Adversarial validation: Challenge results with boundary cases; do not declare convergence until delta <2% over two consecutive runs.
• Logging & recall: Tag MCP/memory artifacts with WHO/WHY to enable longitudinal analysis.

Execution Phases

1. Setup & Scope
   • Choose target skill(s) and datasets; define success metrics and delta thresholds.
   • Prime memory namespace for the session and ingest prior runs.
2. Self-Application Loop
   • Apply the target skill to itself or to curated tasks; record findings, gaps, and fixes.
   • Iterate until improvements plateau; note failures and anti-patterns.
3. Adversarial Probes
   • Inject edge cases, noise, and counterexamples to stress validation paths.
   • Capture false positives/negatives and adjust guardrails.
4. Harvest & Publish
   • Distill reusable patterns, playbooks, and scripts; update references/resources.
   • Summarize deltas, risks, and next steps with confidence ceiling.

Output Format

• Scope, metrics, and target skills for the session.
• Iteration log with deltas, evidence, and remediation actions.
• Adversarial probe results and adjustments.
• Convergence summary, risks, and confidence statement.

Validation Checklist

• Targets and metrics defined; memory namespace prepared.
• At least one self-application and one adversarial probe completed.
• Deltas measured; convergence or stopping condition documented.
• Patterns harvested into references/resources.
• Confidence ceiling included; English-only output.

Confidence: 0.73 (ceiling: inference 0.70) - SOP rewritten to combine Prompt Architect confidence discipline with Skill Forge dogfooding structure.