{ "title": "artgate Phase 3: procedural + code-as-art routes, and the shared finishing stage", "objective": "Run the two CPU-only, download-free cells of the art bake-off - route (a) procedural/parametric and route (d) code-as-art via SVG - against the fixed twelve-subject set, and build the finishing stage that lets any continuous-tone route satisfy a pixel-art spec. Establishes measured cells in the decision table and the shared bridge that route (b) diffusion will reuse.", "acceptance_criteria": [ "The twelve bake-off subjects are pinned in code with their measured easy/hard split, so every route is compared on the same task.", "Route (a) renders all twelve subjects deterministically from a locked palette and its output passes the pixel_icon gate stack.", "Route (d) renders vector subjects via rsvg-convert with no new Python dependency.", "A finishing stage converts continuous-tone output (anti-aliased alpha, thousands of colours) into gate-passing pixel art, demonstrated on real rasteriser output rather than a synthetic stand-in.", "Alpha is binarised BEFORE colour quantisation, mirroring the gate order, so transparent pixels never contribute RGB to the palette.", "100% branch coverage, zero lint suppressions, all pre-commit hooks pass.", "A contact sheet is produced per cell for the human gate." ], "out_of_scope": [ "Route (b) diffusion. The GPU path is proven but no model is downloaded or run in this change.", "The 3D cell of route (d). Blender 5.2 is present and scriptable but not yet wired.", "Deciding which icons are GOOD. The machine gates establish 'not malformed' only; acceptance is the user's contact-sheet verdict.", "Tiling and animation gates (SEAM_ENERGY, LOOP_POPS), which belong to cells not yet run." ], "verifier": "python -m pytest python_pkg/artgate/tests/ --cov=python_pkg.artgate --cov-branch --cov-fail-under=100 (92 tests, 100% branch); pre-commit run over all artgate files; plus an end-to-end run generating all 12 procedural icons and gating them (12/12 machine-pass), and rendering 5 vector subjects before/after the finishing stage (0/5 -> 5/5)." }