# Faultprint Report Output Standard 001

Status: mandatory for all future Faultprint diagnostic reports

Date: July 5, 2026

## Non-Negotiable Rule

The main diagnostic report must carry the proof.

Supporting tables are useful, but they do not satisfy the standard by themselves. A reader must be able to read the report and see exactly where each finding came from.

## SRUOT-Derived Controls Adopted

Faultprint adopts the following report controls from SRUOT v1.4 because they directly strengthen falsifiability:

- No finding without source corpus evidence.
- Each finding must separate source evidence from analyst interpretation.
- Each finding must include a boundary statement naming what the finding does not establish.
- Observations must be derivable from the evidence block.
- Considerations, if used at all, must stay at the class-of-question level and must not become recommendations.
- Scope challenges must be anticipated in the report when a finding could be read more broadly than the evidence permits.
- Null findings matter: if the reviewed corpus was examined for a condition and the evidence was not found, that result must be stated as a bounded null finding rather than ignored.

Faultprint does not adopt SRUOT wholesale. It uses these controls as report-writing guardrails.

## Three-Layer Evidence Architecture

Every Faultprint finding must contain three visible layers:

1. Source Corpus Evidence: the specific document location and evidence excerpt or bounded evidence note.
2. Analytical Reading: the analyst's bridge from the source evidence to the structural finding.
3. Boundary / Falsification: what the finding does not establish and what evidence would weaken, disprove, or reclassify it.

If any layer is missing, the finding is incomplete.

## Required Finding Format

Each finding in the diagnostic report must use this structure:

```text
Finding [ID]: [Finding name]

Finding classification:
[Exposure / Mitigant / Evidence gap / Framework classification]

Evidence Anchor
- Source:
- Corpus locator:
- Evidence excerpt or bounded evidence note:
- Claim type:
- Analytic bridge:
- Confidence:
- Falsification test:
- Boundary:
```

The `Boundary` line is not optional. It replaces vague caveats with a specific statement of what the finding does not establish.

## Findings, Observations, And Open Questions

Faultprint findings must stay inside the evidence perimeter.

- A finding requires source corpus evidence.
- An observation must be derivable from a finding's evidence block.
- An inference without enough source corpus evidence belongs in Open Questions / Evidence Gaps, not in Findings.
- A missing document category can be a finding only when the reviewed corpus was expected to contain that category and the absence is itself structurally meaningful.

## Scope-Calibration Requirement

Every diagnostic report must include a short scope-calibration statement, either inside each finding or in a dedicated section.

The statement must answer:

- What exact corpus supports this finding?
- What broader claim might a reader mistakenly infer?
- What broader claim is not established by the reviewed corpus?

If a finding's wording reaches beyond the evidence base, the finding must be narrowed before delivery.

## Null Finding Requirement

If the diagnostic looked for a material condition and did not find evidence of it, the report should state that as a bounded null finding when it matters to the classification.

Example:

```text
Null finding: The reviewed public corpus did not show recurring harm, failed human review, or post-harm remediation.

Boundary: This does not prove those conditions are absent internally. It only means they were not evidenced in the reviewed corpus.
```

## Pass / Fail Rule

A report fails if:

- any major finding lacks an Evidence Anchor block
- any Evidence Anchor lacks a corpus locator
- any Source Fact lacks a source location
- any inference fails to name the source facts it rests on
- any finding lacks a falsification test
- any finding lacks a boundary statement
- any observation introduces information beyond the evidence block
- any inference without sufficient evidence is placed in Findings rather than Open Questions / Evidence Gaps
- the report makes a recommendation in the standard diagnostic output
- the reader must open a CSV to understand the basis of the finding

## Required QA Artifact

Every run must include:

`06_CLIENT_OUTPUT/report_anchor_audit.csv`

The run is incomplete until every row in that audit passes.