rahmanef63/mso/mso-image-editor

/mso-image-editor — document-level image editing

MSO exposes the same editor command registry to the browser AI and the CLI. The CLI edits document data; the real editor remains the only renderer and owns raster operations.

Resolve paths dynamically:

MSO_ROOT="${MSO_DIR:-$(systemctl show -p WorkingDirectory --value mso.service 2>/dev/null || true)}"
[ -n "$MSO_ROOT" ] || MSO_ROOT="$HOME/mso"
IE="$MSO_ROOT/claude-skills/mso-image-editor/image-editor.sh"
 
"$IE" new 1080x1080
"$IE" open ~/pic.jpg
"$IE" run layer.add kind=text text="SALE" fontSize=160 fill=#ffffff
"$IE" inspect
"$IE" save ~/pic.doc.json
"$IE" view ~/pic.doc.json

Model

• new / open → create a working document session.
• inspect / doc → read document state.
• run → apply one command from the shared registry.
• save → persist a .doc.json host file.
• view → open the persisted document in the real editor for rendering.

Use "$IE" tools as the authoritative command/schema list instead of copying a static command count into this skill.

Boundaries

• Document commands may change layers, text, shapes, transforms, adjustments and styles.
• Brush pixels, interactive masks, background removal and final raster rendering are browser-only by design.
• Every run is stateless apart from the document it returns. Save versions when an undo point matters.
• Never hardcode the MSO repo path or login secret. image-editor.sh resolves its own repo root and reads the configured env file without printing credentials.