Remove canned openings such as “I hope this message finds you well” unless context genuinely calls for them.
Remove inflated AI-style filler such as “delve”, “unlock”, “game-changing”, “seamless”, “robust”, “transformative” when a simpler factual word works.
Avoid excessive em dashes, stacked colon lists, repeated three-part slogans and mechanical transition phrases.
Use one clear reason for contacting the recipient and one primary call to action.
Preserve factual truth boundaries. Never invent traction, customers, revenue, partnerships, endorsements, production scale or urgency.
Preserve the Architect’s approval boundary: editorial changes after approval invalidate that approval and require re-approval.
Preserve privacy: public copy may use Architect, verified corporate email and verified official sites; do not add private location, address, phone or registrant data.
Normalize text to standard Unicode and remove accidental zero-width / bidi control characters before hashing and approval.
The final message should read as deliberate professional correspondence, not as a template or marketing generator output.
Prohibited objective
This skill is NOT an “AI detector bypass” and must not claim that text is human-authored, undetectable, or guaranteed to defeat detection systems. It must not add deceptive typos, false personal anecdotes, fabricated emotions, hidden characters or provenance manipulation.
The acceptable objective is editorial quality: natural professional writing that accurately expresses the Architect’s intended message and is explicitly approved before external transmission.
Quality gate
A draft should pass all of the following before presentation:
recipient-specific context present
concrete purpose present
factual claims supported or clearly framed as architecture/vision
no generic hype cluster
no private-data leakage
one clear CTA
no hidden/control characters
no unapproved changes after approval
Truth boundary: NATURAL_EDITORIAL_STYLE != PROOF_OF_HUMAN_AUTHORSHIP != AI_DETECTOR_EVASION