Adds a prompting section grounded in Tripo3D input guidance and the
text-to-image composition literature (spatial relations, binding,
negation, counting): one body with one analogy that already carries
the layout, multi-body ideas grown as lobes of one mass, thick closed
symmetric forms, matte opaque surfaces, positive frame clause.
Rewrites the three failed prompts as sea turtle / water jar,
manta / beluga and flatiron / hammerhead, keeping the old prompts in
collapsed callouts, and audits the remaining prompts.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ff0779d230