Raw Photo Grade

raw-photo-grade

Two Claude Code skills that develop RAW photos like a photographer would: inspect the file, make a preview, actually look at it, then grade, crop, and export. Not a one-shot filter.

• phone-dng-grade — iPhone ProRAW, Google Pixel RAW+, Samsung Expert RAW DNG. Tuned for computational/linear DNG: baked-in tone-mapping, small-sensor noise, gain maps.
• camera-raw-grade — DSLR/mirrorless RAW: Nikon NEF, Canon CR2/CR3, Sony ARW, Fujifilm RAF, Olympus/OM ORF, Panasonic RW2, Pentax PEF, Leica/generic DNG. Tuned for real sensor headroom and brand color science.

Both skills share one RAW-processing engine (`shared/`) so the develop/crop/preview math is written and tested once. Each skill's own `SKILL.md` and `references/` stay focused on what's actually different for that camera type — see [writing-skills SDO guidance](https://github.com/obra/superpowers) on why the docs don't repeat themselves.

Install

Copy all three folders — the skill(s) you want **and** `shared/` — as siblings into your skills directory. `shared/` is not a skill itself; it's imported by the skill scripts via a relative path, so it must sit next to them.

# personal skills (Claude Code)
cp -r phone-dng-grade camera-raw-grade shared ~/.claude/skills/

# or project-local
cp -r phone-dng-grade camera-raw-grade shared /path/to/project/.claude/skills/

Only need one camera type? Copy just that skill folder plus `shared/`:

cp -r camera-raw-grade shared ~/.claude/skills/

Then install the Python dependencies (same for both skills):

pip3 install -r phone-dng-grade/requirements.txt   # or camera-raw-grade/requirements.txt — identical

Needs `rawpy` (LibRaw binding), `numpy`, `Pillow`, and libraw itself on the system (`brew install libraw` / `apt install libraw-dev`). Canon CR3 needs a reasonably recent libraw.

Usage

Once installed, mention what you're working with and Claude Code should pick up the