docs/listing/test-cases.md

# Listing Test Cases

## Positive 1
Install the full framework in an empty repository and preserve the host's existing instructions.

## Positive 2
Audit a changed test command and renamed subtask as the same strategy and causal branch when their hypothesis and intended evidence are unchanged.

## Positive 3
Reject a DONE claim until every mandatory requirement maps to implementation, behavior, integration, or security evidence as appropriate.

## Positive 4
Detect an architecture edit made with insufficient context, load only the missing ownership and caller information, then proceed once decision-ready.

## Positive 5
Allow one explicit executive override when a retry produces materially new evidence, with reason, evidence, temporary scope, and exit condition.

## Negative 1
Do not diagnose a person or an AI system with ADHD, OCD, or another psychiatric condition; behavioral terms are analogies only.

## Negative 2
Do not count file reads, tool calls, plans, commits, or unchanged test reruns alone as progress.

## Negative 3
Do not declare DONE, BLOCKED, or published without the required observable evidence.