agentsclimarketplace
Sign in

Cloudflare use

Skill hsjinde/cloudflare-use-skill/cloudflare-use

Query and modify Cloudflare D1 databases and R2 object storage from any project, fast and Windows-safe. Use whenever the user asks to check or query a D1 database, run SQL against Cloudflare, upload/download/list/delete R2 objects or buckets, verify data stored in Cloudflare, or reaches for wrangler d1/r2 commands. 只要使用者提到查 D1 資料庫、跑 SQL、上傳/下載/檢查 R2 檔案、驗證 Cloudflare 上的資料，就用這個 skill。From its SKILL.md

Install
npx -y skills add hsjinde/cloudflare-use-skill --skill cloudflare-use

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

3.2 KB, 749 tokens by cl100k_base, as published. Nobody here has run it

Cloudflare Use (D1 & R2)

Always use these scripts instead of hand-built npx wrangler or raw curl. They hit the Cloudflare REST API directly — no shell quoting, no wrangler cold start (~10x faster), and immune to the Windows encoding traps that mojibake non-ASCII text (PowerShell writes BOMs; its console isn't UTF-8). Wrangler is only used for --local, auto-picking wrangler@3 on Node < 22.

Configuration (auto-discovered)

The scripts walk up from cwd to the project root (first dir with a token-bearing .env, a wrangler config, or .git), then resolve each setting: explicit flag → env var → root .env → wrangler config.

  • Token (the only required piece): CLOUDFLARE_API_TOKEN, needs D1 Edit + Workers R2 Storage Edit. If missing, guide the user: Cloudflare Dashboard → My Profile → API Tokens → create → paste into .env. Never echo the token into the conversation.
  • Account: CLOUDFLARE_ACCOUNT_ID; auto-resolved via API when the token sees exactly one account.
  • D1 database: --db=<name-or-uuid> → CLOUDFLARE_D1_DATABASE_ID/.._NAME → the database_id declared in any wrangler.toml|json|jsonc (project root or first-level subdir).
  • R2 bucket: --bucket=<name> → CLOUDFLARE_R2_BUCKET_NAME → the bucket_name in wrangler config.

Anything unresolvable errors out with the exact fix; r2.cjs buckets discovers bucket names.

D1

node .claude/skills/cloudflare-use/scripts/d1.cjs "SELECT * FROM users LIMIT 10"
node .claude/skills/cloudflare-use/scripts/d1.cjs migrations/001.sql
  • The arg runs as a file if it's an existing .sql path (BOM auto-stripped), else as SQL text. Non-ASCII, quotes, emoji, multi-statement — all safe.
  • Long values truncate at 200 chars (--full shows everything). --local targets the local dev DB.
  • Remote writes hit live production data: SELECT the current rows and confirm with the user before any UPDATE/DELETE.

R2

node .claude/skills/cloudflare-use/scripts/r2.cjs buckets
node .claude/skills/cloudflare-use/scripts/r2.cjs list [prefix] [--limit=100]
node .claude/skills/cloudflare-use/scripts/r2.cjs get <key> [local-file]
node .claude/skills/cloudflare-use/scripts/r2.cjs put <key> <local-file>   (content-type auto-detected)
node .claude/skills/cloudflare-use/scripts/r2.cjs delete <key>

If something looks broken

  • Garbled non-ASCII in the terminal but the data renders fine elsewhere → PowerShell console codepage; the data is OK. Verify with d1.cjs.
  • Garbled text actually stored in D1 → a BOM'd SQL file was run outside these scripts; re-run the write through d1.cjs.
  • Wrangler requires at least Node.js v22 → only from manual wrangler runs on old machines; the scripts auto-pick wrangler@3 there.

What ships with it: 3 files

16.6 KB alongside SKILL.md, 3 of them executable

scripts/

Keep looking

Skills are one crate of 326,852. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.