Version: 1.0 Status: Foundational audit specification

The Audit Engine exists to establish what can actually be observed, measured, verified, or reasonably inferred about a defined business/digital subject. It is a diagnostic system, not a sales-score generator.

REQUESTED → SCOPED → COLLECTING → EVIDENCE_REVIEW → FINDINGS → REPORT_READY → DELIVERED. A failed, inaccessible, or insufficiently evidenced audit must be recorded as such rather than represented as completed.

• Observed: directly observed during the audit session or from an explicitly identified source.
• Verified: independently supported by a reliable source or reproducible check.
• Measured: produced by a defined measurement procedure with a known basis.
• Client-provided: supplied by the client and retained as client-provided unless independently verified.
• Estimated: calculated from stated assumptions; never presented as observed fact.
• Target: desired future state, not a current result.
• Unavailable: could not be accessed or measured.
• Inconclusive: available evidence is insufficient for a reliable conclusion.

Every substantive finding must have an evidence reference. Every measurement must identify its measurement method, source, timestamp where relevant, and limitations. A source that cannot be accessed must not be described as scanned or inspected.

A score may be produced only when a documented scoring rubric exists, the required inputs are available, and the calculation is reproducible. Missing inputs may not silently become zero, average, pass, or fail values. If the methodology does not support a score, report Not Scorable and explain why.

The Audit Engine must never fabricate:

• revenue loss or revenue opportunity;
• competitors or competitor activity;
• traffic, conversion, engagement, or performance figures;
• scans, crawls, tests, or observations that did not occur;
• rankings, health scores, grades, or benchmark positions without a valid basis;
• client outcomes or testimonials;
• recommendations presented as measured facts.

Each finding should preserve: finding → evidence → confidence → implication → recommendation → limitation. Recommendations must be distinguishable from observations.

Confidence describes evidentiary strength, not business performance. Use defined labels such as High, Medium, Low, or Inconclusive only when the evidence basis is documented.

An audit must state what was included, excluded, inaccessible, and not assessed. A partial audit must be labelled partial.

Where technically possible, audit checks should record the check definition, input/source, timestamp, and result. A later re-audit must be able to distinguish changed evidence from changed methodology.

Audit Engine: establishes evidence and findings.
Oracle: guides users and explains available information.
Hunter: performs authorized automation.
GROW / CREATIVE / DIGITAL / PROMOTION: act on appropriate findings through governed projects.

Before an audit report is marked ready, the system must verify that substantive claims have evidence references, unsupported measurements are labelled, unavailable checks are disclosed, and any score is reproducible under the declared rubric.

If the system did not observe, measure, verify, or establish it, the audit must not present it as fact.