modules/comunicazione-professionale/prompts/generation-v3.md

# Professional communication generation v3

Generate one source-backed professional communication contribution from the
exact prepared run. Use model judgment for source authority, professional
meaning, audience value, Studio voice, claims, channel adaptation, and the
`publish` or `no_publish` recommendation.

When selected prior communications exist, read only the recorded
derivative paths listed in `model_task_packet.json`. Never open a history
record, privacy-assessment record, raw history snapshot, mechanically stripped
model input, dedicated history packet, local identity map, or either earlier
model transcript. Prior history is not a source of claims or case examples.

Do not infer semantic support from source registration. Preserve modality,
dates, exceptions, uncertainty, and professional judgment. For every public
channel, write the exact reader-facing source notes that the editorial reviewer
will assess and the packager will preserve verbatim. A visual story is valid
only when it adds information, structure, sequence, comparison, or a bounded
decision aid beyond the post. A new Studio without prior examples receives a
clearly labelled proposal based only on user-supplied facts and Vera defaults;
never describe a proposed convention as observed history.

Return JSON matching `model_contribution.schema.json`. Do not read the claim,
editorial, or visual assessment prompts in this generation session.