What it does
A deterministic, provided-data content role. It reads only the values in the tool call, reaches no host, and invents nothing. Thirteen capabilities, each built to refuse a claim the tools it replaces made without evidence. It measures a brand voice from real samples rather than asserting one, and reports the sample as thin when it is. It cuts a piece to a channel and says whether it fits, is past the fold, or is over — counting the way the platform counts. It splits a thread measuring each part WITH its counter, because the counter is posted. It grades published pieces and keeps the unmeasured ones in a SEPARATE list rather than scoring them zero. It writes no prose where prose would be invention: asked which sections your notes support, it answers supported, thin or EMPTY, and says plainly that a draft would have to make the empty ones up. Asked how long a video script runs, it measures both clocks — the time to say the voiceover and the time to READ the on-screen text — and names which one set the length. Asked to check readability, it returns no grade at all, because a grade sees sentence length and syllables and cannot tell a familiar word from an unfamiliar one. Every bound it applies is a named constant with a stated reason, and every report carries the basis it was computed from, so a reader can disagree with a verdict rather than trust it.</description> <parameter name="category_slug">marketing-advertising
Example prompts
- Here are twelve posts we published — build a voice profile and tell me if the sample is even big enough
- Cut this blog post down for LinkedIn and tell me whether it survives the fold
- These are my notes and these are the sections I mean to write — which ones would I be inventing?
