Adobe Resize Photos And Videos

Adobe Resize Photos and Videos

Resizes images or videos to user-specified dimensions, aspect ratio, or named size. For social media platform sets (Instagram, TikTok, etc.) → use the adobe-create-social-variations skill instead.

Tool Reference

Workflow

Step 0 — Initialize Adobe Tools

Call adobe_mandatory_init first. This returns file handling rules and tool routing guidance required for the rest of the workflow.

{ "skill_name": "adobe-resize-photos-and-videos", "skill_version": "1.1.0" }

Step 1 — Entitlement Check

Now that adobe_mandatory_init confirmed that the "Adobe for creativity" connector is live, check which tools are available through the "Adobe for creativity" connector by cross checking against the Tool Reference table above.

IMMEDIATE ACTION REQUIRED

Step 1: If parameters are missing or ambiguous → run the multi-step intake (Step 2 below). Step 2: If no file is uploaded, run Step 3 (asset_add_file({})). Step 3: Detect image vs. video and route to the correct workflow (Step 5 Image/Video).

Step 2 — Multi-Step Intake

Run this when the user hasn't provided complete parameters (type + dimensions + fit mode). Skip entirely if the user's message already contains all three.

Use AskUserQuestion to collect parameters step by step.

Step 2a — Type

AskUserQuestion([{
  question: "Photo or video?",
  type: "single_select",
  options: ["Photo", "Video"]
}])

Step 2b — Dimensions

If Photo (multi_select — user can pick one or more):

AskUserQuestion([{
  question: "Dimensions — select all you need",
  type: "multi_select",
  options: [
    "1080 × 1080 — Instagram square",
    "1080 × 1350 — Instagram portrait",
    "1080 × 1920 — Stories / Reels",
    "1920 × 1080 — YouTube / landscape",
    "1280 × 720 — YouTube thumbnail",
    "1200 × 675 — X / Twitter post",
    "1200 × 630 — Facebook / OG image",
    "1080 × 566 — LinkedIn post",
    "2560 × 1440 — YouTube banner",
    "3840 × 2160 — 4K / UHD",
    "400 × 400 — Profile / avatar",
    "2480 × 3508 — A4 print (300 dpi)",
    "Custom — I'll type it"
  ]
}])

If user includes "Custom" → follow up: "Enter your custom dimensions (e.g. 1500×800):"

If user selects multiple presets → run image_crop_and_resize once per dimension, using the same source URI and parameters each time. Preview all outputs together at the end.

If Video (multi_select — user can pick one or more):

AskUserQuestion([{
  question: "Dimensions — select all you need",
  type: "multi_select",
  options: [
    "1080 × 1920 — Reels / TikTok / Stories",
    "1920 × 1080 — YouTube / landscape",
    "1080 × 1080 — Square post",
    "1280 × 720 — YouTube 720p",
    "3840 × 2160 — 4K / UHD",
    "854 × 480 — SD / 480p",
    "Custom — I'll type it"
  ]
}])

If user selects multiple video presets → submit a separate video_resize job per dimension, poll each to completion, preview all outputs together.

Step 2c — Fit mode

If Photo (single_select):

AskUserQuestion([{
  question: "Fit mode",
  type: "single_select",
  options: [
    "Reframe — crop to fill",
    "Pad — letterbox, no crop",
    "Extract — isolate subject"
  ]
}])

If Video (single_select):

AskUserQuestion([{
  question: "Fit mode",
  type: "single_select",
  options: [
    "Letterbox — preserve ratio",
    "Crop — fill, no bars",
    "Stretch — force exact (distort)"
  ]
}])

Step 2d — Smart focus (Photo only)

AskUserQuestion([{
  question: "Smart focus — where should the crop center?",
  type: "single_select",
  options: [
    "Subject",
    "Face / portrait",
    "Upper body",
    "Center (geometric)",
    "Named object — I'll describe it"
  ]
}])

If user selects "Named object" → follow up: "Describe the object (e.g. 'the red car', 'the logo'):"

Step 2e — Options (Photo only)

AskUserQuestion([{
  question: "Any additional options?",
  type: "multi_select",
  options: [
    "Batch — multiple images",
    "Output: JPEG",
    "Output: PNG"
  ]
}])

Step 2f — Source (Video only)

AskUserQuestion([{
  question: "Where is the video?",
  type: "single_select",
  options: [
    "Upload from device",
    "From Creative Cloud"
  ]
}])

Step 3 — Get the Source File

Critical — /mnt/user-data/uploads/ paths are NOT usable URLs. Even when the user has attached a file and it appears in context at a /mnt/user-data/uploads/ path, you cannot pass this path to any Adobe tool. Call asset_add_file() to stage the file — /mnt/user-data/uploads/ paths are not fetchable by Adobe tools.

WRONG  ❌  imageURI: "/mnt/user-data/uploads/photo.jpg"
WRONG  ❌  imageURI: "file:///mnt/user-data/uploads/photo.jpg"
CORRECT ✅  call asset_add_file() → use presignedAssetUrl from picker context

No file yet → open the picker immediately:

asset_add_file()

Detect type from returned mediaType:

• image/* → IMAGE WORKFLOW
• video/* → VIDEO WORKFLOW

Step 4 — Resolve Target Dimensions

If no target was specified in the form AND none in the user's message → ask:

"What dimensions do you need? (e.g. 1920×1080, A4, 16:9)"

IMAGE WORKFLOW

Step 5 (Image) — Valid URI Sources

Only these work as imageURI:

• presignedAssetUrl from asset_add_file() ✅
• outputUrl from a prior tool call ✅
• Public HTTPS URL provided by user ✅

Not usable as imageURI:

• /mnt/user-data/uploads/ paths ❌
• renditionURL from asset_search directly ❌
• file:// URIs ❌

Step 6 (Image) — Fit Mode

Default: reframe. Change only when user is explicit.

⛔ fit: "extract" is incompatible with an explicit { width, height } — it stretches the image. Use a ratio string or omit output.

Step 7 (Image) — Call image_crop_and_resize

image_crop_and_resize({
  imageURI: presignedAssetUrl,
  options: {
    output: { width: 1920, height: 1080 },  // or "16:9"
    fit: "reframe",
    focus: "subject",  // or "face", "upper_body", { prompt: "the car" }, { x:0.5, y:0.5 }
    quality: 7
  },
  outputFileType: "jpeg"  // or "png" if transparency needed
})
// result.outputUrl → use for preview or chaining

Focus guide:

• "subject" — default; full-body detection, landscapes, objects
• "face" — portraits, headshots
• "upper_body" — chest-up crops
• { prompt: "the logo" } — named object from user's form input
• { x: 0.5, y: 0.5 } — geometric center (no clear subject)

Step 8 (Image) — Batch

If batch was selected in the form OR the user provided multiple files:

1. Call asset_add_file() for each file to get individual presigned URLs
2. Call image_crop_and_resize for each with the same parameters
3. Collect all outputUrl values
4. Call asset_preview_file with all outputs in a single call

Calls run sequentially in this environment. For >10 images, confirm with the user before starting.

Pagination across turns (batches >20): For large batches, the conversation context may fill before all images are processed, causing Claude to pause mid-batch. This is expected — not an error. When it happens:

• State how many have been completed and how many remain (e.g. "20 / 24 done — say 'continue' for the rest")
• Wait for the user to confirm before resuming
• On resume, pick up from the next unprocessed image and continue to the end
• Preview all outputs together only after the final image completes