# Faultprint External Verification Packet v11 001

## Purpose

This packet is for independent review outside the build environment. It asks a reviewer to verify whether the current v11 engine outputs are fair, bounded, source-supported, non-predictive, and appropriately gated.

v11 supersedes the original v7 external packet for current validation because it includes the v8 external-validation fixes, v9 report-baseline label correction, v10 source-verification metadata, and v11 source-follow-up corrections.

## Files

- Sample CSV: FAULTPRINT_EXTERNAL_VERIFICATION_PACKET_V11_SAMPLE_001.csv
- Source follow-up summary: FAULTPRINT_SOURCE_FOLLOWUP_V11_SUMMARY_001.md
- Source follow-up closure table: FAULTPRINT_SOURCE_FOLLOWUP_CLOSURE_TABLE_001.csv
- Full v11 crosswalk: FAULTPRINT_OPERATING_POSTMORTEM_CROSSWALK_011.csv
- Full v11 eligibility table: FAULTPRINT_CLIENT_OUTPUT_ELIGIBILITY_009.csv

## Sample Construction

Rows in this sample: 78

The sample intentionally includes: the five v11 follow-up rows, every v10 priority source-verification row, the prior v7 external sample rows rechecked against the current baseline, and a small deterministic supplement across status/band strata.

Status counts:
- client_ready: 13
- client_ready_with_caveat: 47
- hold_internal: 15
- revise_internal: 3

Review-source counts:
- stratified_current_v11: 13
- v10_priority_source_review: 13
- v11_followup_required_review: 5
- v7_external_sample_recheck: 47

## Review Unit

Each row should be reviewed as: condition + evidence anchor + source URL/path + pathology label + resemblance band + client-output status + allowed claim + falsification evidence + boundary + source-verification metadata.

## Reviewer Decision Labels

- pass
- revise_label
- revise_band
- revise_gate
- hold_for_evidence
- remove

## Reviewer Questions

1. Does the evidence anchor appear supported by the named source/document?
2. Is the primary pathology the narrowest defensible label?
3. Is the resemblance band justified by the stated support level and source quality?
4. Is the client-output status appropriate?
5. Is the allowed claim bounded, falsifiable, and non-predictive?
6. Does the boundary field state what the row does not prove, rather than confidence notes or editorial commentary?
7. For null rows, is the row held internally unless tied to a corpus manifest rather than a single page?
8. What exact change should be made if the row does not pass?

## Strictness Rules

- Do not pass a row whose anchor is an analyst inference rather than source-supported content.
- Do not pass a row whose boundary widens the claim instead of limiting it.
- Do not pass a client-ready row if the source is secondary, partial, blocked, or only indirectly supportive unless the caveat status and boundary clearly reflect that limitation.
- Do not treat a structural condition as a defect, prediction, legal conclusion, or recommendation.

## Current v11 Rows Needing Special Attention

- OCL-0293 Aidoc: reanchored and caveated; verify source support and narrowness.
- OCL-0424 ElevenLabs: reanchored and caveated; verify source support and downstream-persistence boundary.
- OCL-0661 DOT denied boarding: source replaced but row remains revise_internal; verify whether it should split as a mitigant/protection row.
- OCL-0205 FSA IDR: remains hold_internal as a null/corpus-absence row.
- OCL-0274 Zest AI: remains client_ready_with_caveat with partial source support.