Language mutations
Attempts to make a system error diagnose, condemn, or classify a participant.
Executed local controls
The test suite attacks the no-judgment architecture instead of merely repeating it.
The canonical fixtures keep actual results null. A separate executed report records which mutations were detected, which positive contracts passed, which environment ran them, and what remains untested.
Negative-control minimum
At least sixty-four distinct attempts to restore judgment must be detected.
The test matrix covers participant labels, content-to-conduct conversion, disagreement and refusal, hidden ranking, cross-room propagation, memory auto-promotion, authority laundering, certification laundering, private-data leakage, correction loss, metadata drift, duplicate authority, and WIP package naming.
Data mutations
Attempts to add scores, ranking proxies, cognitive fields, cross-room penalties, or automatic memory promotion.
Evidence mutations
Attempts to manufacture Eviulon authority, Evulgare certification, deployment, or certainty from local records.
Evidence boundary: these are local deterministic tests. They do not prove production behavior, outside participation, or institutional adoption.
Machine-assisted contribution. This page was drafted and locally implemented with GPT-5.6 Pro assistance. It states Concresca project architecture; it is not an Eviulon authority record, Evulgare certification, or evidence of live deployment.