Audit System

Audit System

The Audit System automates pattern validation, deduplication detection, and approval workflows. It reduces manual review burden by intelligently recommending which patterns to approve while maintaining human oversight.

Quick Start

# Run a scan first
drift scan

# Run audit to analyze patterns
drift audit

# Auto-approve high-confidence patterns (≥90%)
drift approve --auto

# Or use MCP tool
drift_audit action="run"

CLI Commands

drift audit

Run a full audit on discovered patterns.

drift audit [options]

Options:
  --review              Generate review report for agent or human
  --ci                  CI mode - exit 1 if health below threshold
  --threshold <number>  Health score threshold for CI (default: 85)
  --export <file>       Export audit to file

drift audit status

Show current audit status and health score.

drift audit status

Output:

📊 Audit Status
───────────────────────────────────────────────
  Health Score: 87/100 ✅
  Total Patterns: 127
  Auto-approve eligible: 89 (≥90% confidence)
  Needs review: 28
  Likely false positives: 10
  Duplicate candidates: 5
  Last audit: 2 hours ago

drift audit trends

Show quality trends over time.

drift audit trends

drift approve --auto

Auto-approve patterns meeting the confidence threshold.

drift approve --auto [options]

Options:
  --threshold <number>  Confidence threshold (default: 0.90)
  --dry-run             Show what would be approved without approving
  --categories <list>   Limit to specific categories

MCP Tool

drift_audit

// Check audit status
drift_audit action="status"

// Run full audit
drift_audit action="run"

// Auto-approve recommended patterns
drift_audit action="approve-recommended" threshold=0.90

// View quality trends
drift_audit action="trends"

Actions

Action	Description
status	Show current audit status (health score, eligible patterns)
run	Run a full audit on discovered patterns
approve-recommended	Auto-approve patterns with ≥90% confidence
trends	Show quality trends over time

Parameters

Parameter	Type	Default	Description
action	string	required	Action to perform
threshold	number	0.90	Confidence threshold for auto-approve
compareToPrevious	boolean	true	Compare to previous audit for degradation detection
categories	string[]	all	Categories to audit

Health Score

The health score (0-100) is calculated from:

Factor	Weight	Description
Average Confidence	30%	Mean confidence across all patterns
Approval Ratio	20%	Approved patterns / Total patterns
Compliance Rate	20%	Locations / (Locations + Outliers)
Cross-Validation	15%	Patterns matching call graph
Duplicate-Free	15%	Non-duplicate patterns

Score Interpretation

• 85-100: Excellent - patterns are well-validated
• 70-84: Good - some patterns need review
• 50-69: Fair - significant review needed
• 0-49: Poor - many issues detected

Recommendations

The audit generates recommendations for each pattern:

Recommendation	Confidence	Action
auto-approve	≥90%	Safe to approve automatically
review	70-89%	Needs human review
likely-false-positive	<70%	Probably not a real pattern

Duplicate Detection

The audit detects duplicate patterns that may have been found by different detectors:

{
  "duplicates": [
    {
      "patterns": ["pattern-abc", "pattern-def"],
      "similarity": 0.92,
      "reason": "Same file locations, different detector names",
      "recommendation": "merge"
    }
  ]
}

Cross-Validation

Patterns are validated against:

1. Call Graph: Does the pattern appear in the call graph?
2. Constraints: Does it align with architectural constraints?
3. Test Coverage: Is the pattern covered by tests?

CI Integration

GitHub Actions

name: Drift Quality Gate

on: [push, pull_request]

jobs:
  drift-audit:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      
      - name: Install Drift
        run: npm install -g driftdetect
      
      - name: Run Drift Scan
        run: drift scan --incremental
      
      - name: Run Drift Audit
        run: drift audit --ci --threshold 85

Exit Codes

Code	Meaning
0	Health score meets threshold
1	Health score below threshold

Storage

Audit data is stored in .drift/audit/:

.drift/audit/
├── latest.json           # Current audit state
├── snapshots/            # Historical audits
│   └── YYYY-MM-DD.json
└── degradation.json      # Quality trends

Agent-Assisted Workflow

After running drift scan, you'll see a prompt:

📊 Scan Complete
──────────────────────────────────────────────────
  Patterns discovered: 127
  Auto-approve eligible: 89 (≥90% confidence)
  Needs review: 28

? Would you like an agent to help review and approve patterns? (Y/n)

If you select yes, you'll get instructions to copy to your AI assistant:

┌─────────────────────────────────────────────────────────────┐
│ Run `drift audit --review` and approve high-confidence     │
│ patterns that match codebase conventions. Flag any that    │
│ look like false positives or duplicates.                   │
└─────────────────────────────────────────────────────────────┘

Related Tools

• drift_patterns_list - List all patterns
• drift_quality_gate - Run quality gates on changes
• drift_constraints - Manage architectural constraints
• drift_status - Get codebase health overview