The adapter fetches visual_url from an allowlisted host (https, image types,
10 MB cap, optional read token) and uploads it as multipart sendPhoto, so the
forge can close without breaking post images.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Assistant: claude-code
Assistant-Model: opus
Assistant-Process: 5457@bnt-lap001
Assistant-Session: 792bf84a-3783-4f5d-a811-cded61dd2012
The adapter response now carries telegram_message_id and a t.me/c link.
compose visual resolves a post's image: own visual_url, else the entry's
rendered portrait, else hold an entry with only a visual prompt, else the
HelixForge main image. send-post.sh resolves before touching credentials,
requires an image on public posts and runs the adapter with ATTACH_VISUAL.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Assistant: claude-code
Assistant-Model: opus
Assistant-Process: 5457@bnt-lap001
Assistant-Session: 792bf84a-3783-4f5d-a811-cded61dd2012
Builds internal/adapter and cmd/adapter: the three contracts/r1.openapi.yaml
operations, idempotency on post_id, the reviewed_by/consent-basis refusals,
per-form (R-2/R-3) length limits, and Markdown-to-Telegram rendering, all
against a Sender interface so it's testable without a live bot token.
Every other open task in FT-WP-0001 and FT-WP-0002 needs something this repo
can't supply on its own: the human session bootstrap in FT-WP-0002 T01, the
fluid CLI from fluid-core, or a person's judgment (consent basis, editorial
review, publishing pace). Set both workplans to status: blocked and flagged
FT-WP-0002-T01 needs_human on the hub, since it's the single blocker gating
everything else.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Assistant: claude-code
Assistant-Model: sonnet
Assistant-Process: 278411@bnt-lap001
Assistant-Session: 19568045-15bc-4b88-ba3d-8e95b4f9c6f1
