Cortex

Cortex is an authorization-context analyzer. It scores incoming requests against a 14-sample reference corpus spanning malware, supply-chain attacks, ICS, social engineering, and LLM injection, and returns a structured “authorization premise” rather than a binary allow/deny.

Why this exists

Most tooling treats authorization as a yes/no gate. In practice, the premise of an authorization claim, who’s asking, on whose behalf, with what stated scope, and what evidence, carries more signal than the yes/no answer. Cortex makes the premise inspectable.

What it does

• 14-sample reference corpus of authorization patterns
• Structured premise extraction
• Confidence scoring per claim
• Pluggable into VisorRAG to gate which probes get considered

Source

github.com/nuclide-research/cortex-framework