Skip to main content

Component: Provenance/Taint Engine

Design-stage

Phase 4, specified in docs/phases/phase-4-transaction-tokens-taint-identity.md and specs/004-transaction-tokens-taint-identity/ but not yet implemented — no TAINTED_INTO_PRIVILEGED deny reason or taint field exists on the PDP or audit rows in this codebase today.

What it is (designed): a label tracker that tags every result with an origin trust label and propagates that label to whatever consumes it downstream.

What it does (designed): blocks a call from tainted content into a privileged target before dispatch — structurally, not as an advisory warning.

Why this is a PDP-level check, not a separate content scanner: enforcing taint in the same decision that already checks scope, edge, and hop budget means it inherits the same commit-before-respond audit guarantee — there is exactly one place, not two systems that could disagree, where "can this call happen" is decided. See The taint and provenance model for the full rationale, including why transaction tokens are this model's companion control.