Getting the archives

The WACZ files are stored with Git LFS. Three ways to obtain them:

# A. Clone (with git-lfs installed, the real bytes are smudged in)
git clone https://github.com/uraitakahito/wacz-validator-corpus

# B. Fetch one over HTTP — from the media URL; the raw URL will not do
base=https://media.githubusercontent.com/media/uraitakahito/wacz-validator-corpus/main/fixtures
curl -sL "$base/good.wacz" -o good.wacz

# C. Regenerate deterministically from wacz-validator — no download, byte-identical
#    CORPUS_DIR must be absolute: --filter runs the script in packages/core,
#    so a relative path resolves there and quietly finds nothing.
CORPUS_DIR="$(cd ../wacz-validator-corpus && pwd)" pnpm --filter @wacz-validator/core corpus:build

Option C exists because the specimens are built, not collected: the same inputs produce the same bytes, so a corpus checkout is a convenience rather than a source of truth.