SOLID Principles Audited Regularly

Severity: Recommended

Code should follow SOLID principles (Single Responsibility, Open-Closed, Liskov Substitution, Interface Segregation, Dependency Inversion) and be regularly audited for compliance. AI agents are well-suited to this assessment.

Check automatically:

1. Check for AI code review integration:

# Check for Claude/AI agent configuration
ls -la CLAUDE.md AGENTS.md .claude* 2>/dev/null

# Check for AI reviewer in GitHub Actions
grep -rE "claude|anthropic|openai|ai-review|code-review" .github/workflows/ 2>/dev/null

# Check for architecture guidelines in agent config
grep -iE "solid|architecture|single responsibility|dependency injection" CLAUDE.md AGENTS.md 2>/dev/null

2. Check for architecture documentation:

# Look for architecture decision records (ADRs)
find . -type d -name "adr*" -o -name "decisions" 2>/dev/null
find . -type f -name "*.md" | xargs grep -liE "architecture|solid|design principle" 2>/dev/null | head -10

# Check for documented coding standards
ls -la CONTRIBUTING.md docs/architecture* docs/coding-standards* 2>/dev/null

3. Proxy metrics (heuristics, not definitive):

# Very large files may indicate SRP violations (>500 lines)
find src -name "*.ts" -o -name "*.js" -o -name "*.py" 2>/dev/null | xargs wc -l 2>/dev/null | sort -n | tail -20

# Files with many imports may have coupling issues (>20 imports)
for f in $(find src -name "*.ts" 2>/dev/null | head -20); do
  count=$(grep -c "^import" "$f" 2>/dev/null || echo 0)
  if [ "$count" -gt 20 ]; then echo "$f: $count imports"; fi
done

# God classes - files with many exported functions/classes
grep -rE "^export (function|class|const)" src --include="*.ts" 2>/dev/null | cut -d: -f1 | sort | uniq -c | sort -n | tail -10

AI agent assessment (the core verification):

• AI agent should review codebase for SOLID violations during PR review
• Periodic architecture audits (monthly/quarterly) by AI agent
• Agent checks for: god classes, interface bloat, concrete dependencies, inheritance misuse
• Agent flags issues with specific recommendations

Ask user:

• "Is AI code review part of your PR process?"
• "How are architecture decisions reviewed?"
• "When was the last architecture audit?"

Cross-reference with:

• ARCH-002 (code complexity - related health metric)
• TEST-005 (CRAP score)
• FLOW-002 (AI + human review in PR process)

Pass criteria:

• AI agent is part of code review process and checks for architecture issues
• Architecture principles documented (in CLAUDE.md, ADRs, or similar)
• No major SOLID violations in critical paths (as assessed by AI review)
• Evidence of recent architecture review

Fail criteria:

• No AI-assisted code review
• No documented architecture standards
• Known god classes or tightly coupled code ignored
• "We don't do architecture reviews"

Evidence to capture:

• How SOLID compliance is assessed (AI agent, manual review, both)
• Location of architecture guidelines
• Recent examples of AI catching architecture issues
• Date of last architecture audit

---