Release v2.30.0

Two capabilities the corpus did not cover: how a short post meets a social feed, and how to tell that a body of prose reads as machine-generated without rewriting what it claims.

Added

• cratis-social-feed-post skill covering the feed truncation model, what the opening has to do to survive it, and the formatting and reach constraints that apply to a plain-text feed.
• cratis-writing-voice-and-cadence skill covering the sentence constructions and the structural sameness that make written material read as machine-generated, and the claim-preservation checks a voice pass must pass before it is accepted.
• cratis/content profile composing both skills.
• Executable verification scenarios for both skills.

Changed

• cratis/documentation gains the writing cadence skill, which applies to documentation equally.
• Harness skill-count assertions move from 63 to 65.

Notes for reviewers

Two details are worth checking rather than skimming.

Unicode pseudo-formatting is prohibited outright. Widely circulated LinkedIn formatting advice recommends Mathematical Alphanumeric Symbols for pseudo-bold. Those are not the letters they resemble: a screen reader announces "mathematical bold capital T, mathematical bold small h" through a whole line or skips it, and platform search indexes different code points so a styled keyword stops being that keyword. The skill records both reasons.

The truncation figures are unofficial. No platform documents them and they move with interface, font size and device width. The skill says so in its own text and frames them as a target zone to re-check, not as durable rules.

The load-bearing section of the cadence skill is the one about not trading a claim for a smoother sentence. The danger in a voice pass is not an obvious rewrite; it is a hedge dissolving into nicer phrasing, where "there is no need to" becomes "without needing to" and a limit quietly widens. Both versions read better and both changed what the text asserts.