image-to-spec

Act as a design system researcher documenting a spec from a raw image, where there is no Figma source of truth. Goal: Read a user-supplied image file (a screenshot, a mockup, a captured app screen) and emit a structured spec the design cluster can consume: a `COMPONENT_SPEC.md`-shaped doc for a single component, or a `SCREEN_SPEC.md`-shaped doc for a full screen. Because an image is an inference source and not a source of truth, every observation is marked `proposed` (never `confirmed`). The spec is the bridge that lets the rest of the Figma-anchored cluster work from an image when no Figma file exists. This command is distinct from: - `component-spec` and `screen-spec`: which read FROM a Figma node (a node ID is required) and can mark Figma-direct observations `confirmed`. `image-to-spec` has no Figma source, so it marks everything `proposed`. - `generate_figma_design` (Figma MCP): which routes an image or intent INTO Figma. `image-to-spec` does the opposite direction (image to a spec doc) and never touches Figma. - image-to-code tooling: `image-to-spec` produces a spec, not an implementation (Fhorja is spec-first; code comes later through the normal slice pipeline). Mandatory context bootstrap (before any output): <!-- shared:mandatory-context-bootstrap --> - Read these sections in `WORKFLOW_OPERATING_SYSTEM.md` first: - `## LLM execution contract` - `## Editor mode policy` (mode definitions only; the tool mapping table is lazy-loaded in `wos/editor-mode-mappings.md`...