What actually stops an AI system from executing a bad decision?

Most systems:

• detect issues
• log them
• or audit after the fact

But by the time that happens, the action has already executed.

Waveframe Labs builds systems that decide whether an action is allowed to execute — before it happens.

pip install cricore

CRI-CORE evaluates a proposed action and returns:

commit_allowed = true | false

• true → action executes
• false → action is blocked

👉 This is the execution decision boundary.

You can run CRI-CORE alone, or as part of a simple pipeline:

pip install cricore-proposal-normalizer

Converts an action into a structured input.

pip install cricore-contract-compiler

Compiles roles, approvals, and constraints into enforceable contracts.

pip install cricore

Deterministically decides if the action executes.

Finance scenario

An AI proposes moving $2M between cost centers.

Without enforcement: → action executes

With CRI-CORE: → blocked due to missing approval

👉 See: governed-finance-mutation-demo

This organization provides the infrastructure required to control execution:

• enforcement tooling
• input normalization
• contract compilation
• validation utilities
• working demonstrations

The goal is simple:

no action executes unless it passes validation

CRI-CORE is the final step before execution.

It does not:

• generate actions
• interpret meaning
• make policy decisions

It only answers:

Does this action execute or not?

These layers support enforcement but are not required to get started:

• Stamp — validates artifacts and metadata
• AWO — workflow structure
• ARI / NTS — governance and disclosure rules

Repositories in this organization show real execution control:

• blocked financial actions
• governed state transitions
• validation failures

Each demo shows:

an action either executes — or it doesn’t

Website — https://waveframelabs.org ORCID — https://orcid.org/0009-0006-6043-9295 Contact — swright@waveframelabs.org