articulate is the conductor for writing: it matches ceremony to the piece, picks the platform up front, and drives the pipeline to a finished, platform-ready draft — never reimplementing a phase inline when a mode owns it. Its sibling materialize owns code; articulate owns prose.

Pick the platform first

A blog post and a LinkedIn post differ from the first fragment, not just at the end — length, hook, what's worth mining, voice. So choose the target platform before mining, and let it tune the whole run. Load its profile from reference/platforms/ (linkedin.md, blog.md, twitter.md, …) and keep it in context across every phase. If the user hasn't said, ask which platform before starting.

Adding a platform is a new profile file, not a new mode. Repurposing a finished piece for a different platform (a blog → a LinkedIn teaser) is a fresh, platform-committed run seeded from the first.

Setup

If the user invoked a mode (init, fragments, shape, beats), read reference/<mode>/<mode>.md next.
Voice profile. Load docs/writing/author.md — the author's voice, differentiators, hard rules, and cadence — and keep it in context across every phase; it's what makes the writing theirs. If it's missing, offer to run init before the first real piece (a writing session works without it, just more generic).
Load the target platform's profile (above). Voice (author) × format (platform) combine on every phase.
Read a sample of the author's existing writing on that platform when available, to match their voice before producing anything.

The pipeline

Run it once; scale the ceremony to the piece. Every phase is platform-aware — it reads the profile.