Internal Audit Nonconformity Determination

Problem

During internal audits of management systems, nonconformity findings are often tracked in spreadsheets or meeting notes, with classification left to individual auditor judgment. Verification status scatters across sessions, and end-of-audit summaries are assembled ad hoc. This skill pins the full NC lifecycle into a local script-plus-JSON workflow, reducing manual data shuffling and inconsistent grading.

How It Works

• Intake & Query: add_nc.py accepts --title, --clause, --description, --evidence, --category, writes to assets/nc_data.json, and returns a unique ID; list_nc.py filters by --status, --clause, --audit-id, etc.
• Rule-Assisted Evaluation: evaluate_nc.py references references/iso_clauses.md to classify input text into major / minor / observation, returning a suggested category and rationale for the auditor to review rather than replace human judgment.
• Verification Tracking: verify_nc.py enforces the open → in_progress → verified → closed state machine and records --verifier and --result.
• Report Generation: generate_report.py aggregates all NCs under a given --audit-id; --stats-only limits output to numeric summaries.

Boundaries

• Requires only the Python standard library (json, argparse, datetime); no third-party packages.
• Clause numbers must match the specific ISO standard revision in use; references/iso_clauses.md covers common clauses but not every version.
• Provide at least two pieces of evidence per finding; ensure all NCs share the same audit ID before generating reports, or statistics will be incomplete.
• The rule engine is assistive; final severity is decided by the audit team at the closing meeting.