How to use the Tigris Object Storage CLI agent skill

What an agent skill is and what this one does

An agent skill is a folder containing a SKILL.md file and any supporting scripts. When it is installed, an AI agent that supports skills can read the instructions and run the commands they describe. This skill's SKILL.md tells the agent to call the official tigris CLI through a wrapper named cli-anything-tigris.

• List and create buckets, and get bucket info
• Upload, download, copy, list, inspect and delete objects
• Create presigned download and upload URLs that expire after a set number of seconds
• Take point-in-time snapshots of a bucket and list them
• Create, scope, rotate and delete access keys
• Check which account or organization you are logged in as, and list or create IAM policies and org users

Install the Tigris CLI, log in and install the skill

The skill needs three things: the official Tigris CLI on your PATH, a logged-in Tigris account, and the Python package. The source lists two ways to install the Tigris CLI: npm or Homebrew. Python 3.10 or newer is required.

There are two ways to add the skill itself. You can download the skill ZIP from the page where you found this guide, then unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder to limit it to that project. You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.

• Install the Tigris CLI with npm install -g @tigrisdata/cli, or with brew install tigrisdata/tap/tigris
• Log in with tigris login, which opens a browser OAuth flow
• Install the Python harness with pip install cli-anything-tigris
• Confirm access with cli-anything-tigris --json auth whoami

Trigger the skill with natural requests

You do not need to type commands yourself. Once the skill is installed, ask the agent for the outcome you want, and name Tigris or the bucket so the agent knows which storage to use. The agent then chooses the matching cli-anything-tigris command.

The skill is built to return JSON from every command, so the agent can read results reliably. Where the skill's guidance applies, the agent is expected to check return codes and read stderr when something fails.

• "List the buckets in my Tigris account."
• "Upload ./report.pdf to bucket reports under the key q3/report.pdf."
• "Make a link that expires in one hour for q3/report.pdf in bucket reports."
• "Take a snapshot of bucket reports named before-cleanup, then tell me what is in the bucket."

Run the commands directly

You can also run the CLI yourself to check what the agent will do. Every command accepts --json for machine-readable output. Running the CLI without a subcommand opens an interactive REPL with tab-completion and history.

Object commands accept t3:// or tigris:// URIs for server-side copies. The source notes that copies from t3 to t3 skip the round trip through your machine.

• Basic help: cli-anything-tigris --help
• Upload: object put --bucket B --key K --file F
• Download: object get --bucket B --key K --output F
• Presigned download link valid for 3600 seconds

Keep destructive work safe with snapshots and scoped keys

Bucket deletion, access-key deletion and access-key rotation all require an explicit --yes flag. The skill's guidance for destructive work is to take a snapshot of the target bucket first, do the work, then keep or discard the snapshot.

For per-agent isolation, the source suggests creating a dedicated access key for each agent run, assigning it to one bucket with a role such as Editor, and deleting the key when the run ends. Access keys are shown only once at creation, so store the secret when it appears.

• Snapshot first: snapshot take my-bucket --name baseline-v1
• Create a key: access-key create my-agent-key
• Scope it: access-key assign tid_AaBb --bucket my-bucket --role Editor
• Revoke when done: access-key delete tid_AaBb --yes

Safety notes: scripts, network and permissions

This skill runs scripts and uses the network. It shells out to the tigris binary and talks to Tigris on your behalf, so anything the agent does with your login happens with your account's permissions. Review the commands an agent proposes, especially deletes, rotations and policy or user changes.

The skill is for Tigris only. It does not manage AWS S3, Cloudflare R2, MinIO or other S3-compatible endpoints.

• Give the agent a scoped key rather than your own login where possible
• Require the agent to take a snapshot before any delete or rotate
• Watch for --yes in proposed commands and confirm each one
• Do not use this skill for non-Tigris storage