Auditor Assurance Pack

Audience: anyone who tests whether controls work: internal audit, HSE audit teams, certification and regulatory inspectors, insurers. They need testable criteria and evidence to request, not narrative.

Engine: PMLC = python3 <skills-root>/learnings-coach/scripts/pmlc.py (this skill needs the learnings-coach skill installed alongside it). Read learnings-coach/references/learning-principles.md and the audit test point block in output-style.md.

Gather

PMLC select <ws> --level 3 4 --top-rated --brief
PMLC select <ws> --tag assurance --brief                    # how existing audits performed
PMLC select <ws> --text "record|log|register|sign|closure|overdue|schedule|audit" --brief

Write

1. Why existing assurance did not catch it: what the findings say about audits, inspections and records before the event (generic checklists, open recommendations, closures without data). Factual, no blame.
2. Audit test points: twelve to twenty, grouped by management-system element (maintenance and testing, change management, materials and spares, permits, competence, emergency preparedness, records, audit follow-up). Each: expectation → what to sample (and how many) → evidence of conformity → red flags seen in this event → evidence trail.
3. Cross-site verification protocol: a short plan to test whether the top five conditions exist at other sites (sampling approach, records to pull, interviews by role, a site visit walk-path).
4. Evidence request list (interviews follow "Interviews in a just culture" in learnings-coach/references/learning-principles.md): the documents and records an auditor should request, by function.
5. Reporting: a grading scale (conforming / partly / not / not tested) and a results table template.

Output

Write the pack <run>/outputs/auditor-assurance-pack.json (schema: learnings-coach/references/pack-schema.md), then:

PMLC check-pack <ws> <run>/outputs/auditor-assurance-pack.json   # fix every problem it lists
PMLC pack <ws> <run>/outputs/auditor-assurance-pack.json         # -> auditor-assurance-pack.html (primary) and .md

Sections → kinds:

1. para.
2. test_point ×12–20 (requirement, sample, conformity_evidence, red_flags; must cite), with group = the management-system element.
3. table with columns Condition, Records to pull, Interview and Walk-path stop, plus row_cites.
4. check items, with group = the function.
5. A para giving the grading scale, then a table results template with empty Grade and Notes cells.

Set "scope": {"levels": [3, 4], "min_rating": "top"}. Put data-quality corrections in data_quality, findings left out on purpose in exclusions, and same-condition findings in supporting_cites. If asked for an .xlsx, build it from the test_point items (one row each) with the host's spreadsheet tools.

{"kind": "test_point", "group": "Maintenance and testing", "requirement": "…", "sample": "…",
 "conformity_evidence": "…", "red_flags": ["…"], "cites": ["FND-…", "EVD-…"]}