A Claude Code skill that teaches the agent s5cmd — blazing-fast, massively parallel S3 data movement — so it reaches for the right command instead of slow aws s3 loops.

• ⚡ Picks the fast path — nudges Claude toward s5cmd (10–30× faster than aws s3) for bulk S3 transfers instead of boto3 loops
• 🎯 Grounded in the binary — every command and flag is lifted straight from s5cmd --help (v2.3.0), not model memory
• 🌀 Full surface — all subcommands: ls cp mv rm mb rb sync du cat pipe select run head presign bucket-version
• 🪤 Knows the traps — wildcard quoting, global-vs-command flag placement, --numworkers vs --concurrency, sync --delete previews
• 🔌 Endpoint-aware — MinIO, Ceph, Cloudflare R2, GCS, Backblaze B2, Wasabi, and other S3-compatible stores
• ✅ Tested — validated with a retrieval scenario before shipping (per the writing-skills TDD method)

Claude loads the skill on S3 data-movement concerns: copying, uploading, downloading, moving, syncing, listing, or deleting objects; transfers that are too slow via aws s3 / boto3; batching thousands of operations; streaming with cat/pipe; SQL select over CSV/JSON/Parquet; presigned URLs; object metadata (head); storage usage (du); bucket management — or any mention of s5cmd itself.

The skill is a single SKILL.md. Drop it where your agent discovers skills:

git clone https://github.com/4thel00z/s5cmd-skill ~/skills/s5cmd-skill
ln -sfn ~/skills/s5cmd-skill ~/.claude/skills/s5cmd

You also need the s5cmd binary on PATH:

brew install peak/tap/s5cmd          # macOS
# or: go install github.com/peak/s5cmd/v2@master
s5cmd version                         # expect v2.3.0+

s5cmd-skill/
  SKILL.md     # the skill: frontmatter triggers + command reference
  README.md    # this file

# Download a whole prefix from a custom endpoint, fast (always quote the glob!)
s5cmd --endpoint-url https://minio.lan:9000 cp "s3://data/raw/*" ./raw/

# Mirror with deletions — preview first
s5cmd --dry-run sync --delete ./out/ s3://backups/

# Delete 50k keys from a list, in parallel
cat keys.txt | awk '{print "rm s3://bucket/" $0}' | s5cmd run

# Stream a download straight into S3, no local disk
curl -sL https://example.com/file | s5cmd pipe s3://bucket/file

MIT

_{Teaches the agent to move bytes at wire speed. 🚀}