add-photos

name	add-photos
description	Add new photos to the wrangell.dog repo — rename to dog-N.jpeg convention, generate captions via vision, update captions.js. Use when the user says "add photos", "I added images", "new photos", or when untracked image files appear in public/images/.
metadata	{"scope":"project","category":"content"}

Add Photos

Handles the full photo intake workflow: rename → caption → update captions.js → stage for commit.

What this skill does

1. Finds any untracked or newly-added image files in public/images/ that don't follow the dog-N.jpeg naming convention
2. Renames them sequentially from the next available number
3. Views each image directly and writes a caption in the established style
4. Appends entries to captions.js and updates the header comment
5. Stages everything — images + updated captions.js — ready to commit

Step 1 — Find and rename

Run git status to find untracked files in public/images/. List existing dog-N.jpeg files to find the highest N.

Rename each new file to dog-<next>.jpeg in order. The order doesn't matter for site correctness (the homepage rotation is date-seeded and any order works), so process in whatever order git status lists them.

Step 2 — Generate captions

Read each new image directly with the Read tool (you can view images). For each one, write a caption following this style:

Caption rules:

• 2–5 words
• No articles (a, an, the)
• Wry and observational — describe what's actually happening, not what you wish were happening
• Must read standalone — no "again", "also", "same", or any word that implies context from another photo
• Present tense participial phrases or compressed noun phrases work best
• One strong specific detail beats a generic description

Examples of the style:

• "Ice cream, overhead" (not "Dog balancing ice cream sandwich on head")
• "His bed now" (not "A man sharing the dog bed")
• "Two tongues, one autumn" (not "Two happy dogs outside in fall")
• "Teeth, briefly visible" (not "Someone squeezing a dog's snout")
• "Turbaned, briefly" (not "Dog with a towel on its head")

After generating all captions, present them to the user as a numbered list with the image filename before asking to proceed. This lets the user tweak any before they're committed.

Step 3 — Update captions.js

public/captions.js structure:

// Captions for dog-1.jpeg .. dog-N.jpeg (index 0 = dog-1). Auto-generated by scripts/caption-new-images.js.
window.WRANGELL_CAPTIONS = [
  "Caption for dog-1",
  ...
  "Caption for dog-N"
];

• Append new caption strings to the array (one per line, comma-separated, matching existing formatting)
• Update the header comment to reflect the new last image number
• The array index is 0-based: captions[0] = dog-1, captions[N-1] = dog-N

Step 4 — Stage

git add public/images/dog-*.jpeg public/captions.js

Report what was added: N new photos (dog-X through dog-Y), N captions written. Leave the commit to the user unless they ask you to commit.

How the photos are used

Understanding these keeps you from breaking things:

Gallery (gallery.html): Uses captions.length as the total plate count. Generates images/dog-${n}.jpeg for n = 1..total. Adding a photo without a caption means the gallery will try to load an image that doesn't exist at that index — always keep images and captions in sync.

Homepage (index.html): Picks today's photo via (year*10000 + month*100 + day) % captions.length + 1. Every time photos are added, the rotation shifts for all future dates. That's expected and fine.

Service worker (sw.js): captions.js is in the app shell and served network-first. Images are cache-first (immutable). No action needed here.

Pre-commit hook (.githooks/pre-commit): Runs scripts/caption-new-images.js, which calls the Claude API if ANTHROPIC_API_KEY is set. Since you're handling captions manually in this skill, the hook will report "captions: up to date" and pass cleanly without the key.

Edge cases

• If the user drops files that are already named dog-N.jpeg correctly, skip renaming — just check they have captions.
• If a photo already has a caption entry (e.g., the array was manually extended), skip it.
• If an image number would create a gap (e.g., dog-103 exists but dog-104 doesn't and the user dropped dog-106), fill sequentially from the first gap — never skip numbers.
• HEIC, PNG, or other formats: rename to .jpeg extension only if the file is actually JPEG-encoded; otherwise note to the user that the file needs conversion first. Most iPhone exports are already JPEG regardless of original extension.