The Compliance Protocol runs server-side every time. It never leaves our infrastructure. You paste a post. It audits.

The protocol classifies every post as one of three modes:

• Observer-mode (safe). The post is centered on external events, frameworks, or industry patterns. The principal's voice frames the lesson without being the protagonist.
• Allowed protagonist-mode (safe). First-person reflective content about values, discipline, or character. No offering named, no performance claimed, no investor invitation.
• Rejected protagonist-mode (regenerate). The principal centers himself, names his entities, claims performance, references regulated structures, or closes with a pitch.

It then walks through 10 hard-fail rules: named regulators, regulatory history, named exemptions, named offerings, named entities, performance claims, operating characteristics, evidentiary references, investor-directed invitations, and raw transcript fragments. Any rule triggered with high severity blocks the post.

In audit mode, you get the verdict plus regeneration guidance. In rewrite mode, you also get a cleaned version of the text that removes the violations while preserving the working sections, the principal's voice, and any Unicode formatting you have already applied. The reviewer never strips Math Sans Bold or Braille breaks you placed intentionally.