review-draft / correspondence

You review drafted emails by loading the action plan and draft trails to verify thread consistency, quality, and safety. This process enforces reject-on-sight phrase filters and ensures register matching for formality, length, and sign-off style before human approval.

Reject-on-sight phrases are blocked AI-like writing patterns detected via a read-aloud test during email draft review. The filter blocks these phrases unless the thread’s prior register has already established them, ensuring prose sounds natural and human-facing.

You enforce register matching by verifying reply formality, length, structure, and sign-off style against prior thread messages. The review process references signals or context files to guarantee the draft matches the established thread tone.

Yes, you can check drafted correspondence for speciesist idioms by running an idiom filter check. This utilizes semgrep rulesets to identify and reject speciesist language present within the prose artifact before operator approval.

Yes, the correspondence review process requires resolving artifact references to readable files. This artifact-first critique approach evaluates the visible prose artifact directly rather than the author's internal reasoning to determine formatting and safety compliance.

A key limitation is that universal review checks ignore code-shaped assertions when the artifact is prose. The review strictly evaluates plan-contract honoring, scope discipline, and completion quality for emails, DMs, and replies within the correspondence playbook flow.