version: "1.0.0" name: social-resizer description: >- Process a directory of images for Instagram using sharp (resize/crop/pad) and pilgram (Instagram filters). Reads a config.json and writes processed images to an output directory. metadata: { "openclaw": { "emoji": "📸", "requires": { "bins": ["node", "python3", "uv"] }, }, }

filter — Instagram Image Pipeline

Two-stage pipeline that prepares images for Instagram:

• Stage 1 — resize, crop, or pad to exact Instagram dimensions (Node.js / sharp)
• Stage 2 — apply an Instagram-style color filter (Python / pilgram)

The skill directory (where this SKILL.md lives) is referred to as $SKILL_DIR below.

When to Use

Use this skill when the user wants to:

• Prepare photos for Instagram (resize, aspect ratio, padding)
• Apply Instagram-style filters to images (clarendon, gingham, etc.)
• Batch process a folder of images for social media

Setup (first run only)

Install dependencies from the lock file — this guarantees reproducible versions:

# Python deps (locked)
cd "$SKILL_DIR/scripts" && uv sync
 
# Node deps (locked via package-lock.json)
cd "$SKILL_DIR/scripts" && npm install

Both lockfiles (uv.lock, package-lock.json) are committed to the skill repo — no version drift.

Agent Workflow

When the user invokes this skill, follow these steps:

1. Ask the user for their intent before doing anything

Present the key options and confirm before writing any config or running the pipeline:

Before I run the pipeline, here's what I need to know:
 
📐 Format
  - portrait  (1080×1350, 4:5) — most common for feed photos
  - square    (1080×1080, 1:1)
  - landscape (1080×566, 1.91:1)
  - custom    (you specify width and height)
 
✂️ Fit (how to fill the canvas)
  - cover   — crop to fill (no padding, may cut edges)
  - contain — pad with a background color (no cropping)
  - fill    — stretch to fit (may distort)
 
🎨 Filter (optional)
  Available: _1977, aden, brannan, brooklyn, clarendon, earlybird, gingham,
  hudson, inkwell, kelvin, lark, lofi, maven, mayfair, moon, nashville,
  perpetua, reyes, rise, slumber, stinson, toaster, valencia, walden, willow, xpro2
  Or: none (skip filter, geometry only)
 
📁 Input / output directories
  Default: ./input and ./output
 
Which settings would you like?

Wait for user response before proceeding.

2. Write config.json

Based on the user's answers, write a config.json in the user's working directory:

{
  "format": "portrait",
  "custom_dimensions": null,
  "background_color": "#ffffff",
  "fit": "cover",
  "filter": "clarendon",
  "output_format": "jpeg",
  "quality": 90,
  "input_dir": "./input",
  "output_dir": "./output"
}

See the Config Reference section below for all options.

Multiple configs: If the user wants to try several filters or formats, write multiple config files (e.g. config-clarendon.json, config-gingham.json) and run the pipeline once per config. Output dirs should be distinct so results don't overwrite each other.

3. Place images

Ensure the user's images are in input_dir. If they haven't copied them yet, tell them to do so and wait, or copy them yourself if you have the paths.

4. Run the pipeline

cd <user-project-dir>
uv run --project "$SKILL_DIR/scripts" python "$SKILL_DIR/scripts/process.py" --config config.json

For multiple configs:

for config in config-clarendon.json config-gingham.json config-lark.json; do
  uv run --project "$SKILL_DIR/scripts" python "$SKILL_DIR/scripts/process.py" --config "$config"
done

5. Report results

After the run, tell the user:

• How many images succeeded / failed
• Where the output files are
• If any errors occurred, show the error messages

Config Reference

Supported input formats: .jpg, .jpeg, .png, .webp

Error Handling

• Missing input_dir → pipeline exits with a clear error before touching any files
• Invalid config values → all validation errors are printed before any images are processed
• Per-image failures → reported individually; other images continue processing
• Non-zero exit code if any image failed — check the summary line at the end of output