akt CLI Agent Skill

The akt-cli skill teaches a coding agent to operate an installed akt binary: inspect contexts, write SDL, deploy workloads, check leases, and recover from failed operations. It includes the main instructions, three reference guides, and optional agent metadata.

Get the Skill

Download the complete akt-cli skill ZIP, or read the files hosted on this website:

The ZIP contains an akt-cli/ directory with all six files. Keep the folder intact so the relative links in SKILL.md and its references continue to work. A SHA-256 checksum is available for the ZIP.

Install the CLI

The skill needs akt on your agent’s PATH and access to a shell. On macOS or Linux with Homebrew:

brew tap akash-network/tap
brew update
brew install akash-network/tap/akt
akt version --long

Homebrew installs the latest published CLI release available in the official tap. For an existing installation, run brew update followed by brew upgrade akash-network/tap/akt. Check GitHub for the most up-to-date code and GitHub Releases for published builds. See Installation for other platforms and source builds.

Install the skill separately using the steps below. You can inspect contexts and validate SDL before configuring credentials; deploying also requires the appropriate Console account or funded local wallet.

Install the Skill

Codex: This Project

From your project root, download and extract the bundle into .agents/skills/:

curl -fL https://akash.network/skills/akt-cli.zip -o akt-cli.zip
mkdir -p .agents/skills
unzip -n akt-cli.zip -d .agents/skills

The resulting entry point is .agents/skills/akt-cli/SKILL.md.

Codex: All Your Projects

To make the skill available across your projects, extract the downloaded ZIP into your user skills directory instead:

mkdir -p ~/.agents/skills
unzip -n akt-cli.zip -d ~/.agents/skills

Choose either the project or user location to avoid duplicate skill entries. These commands preserve existing files. To update an older copy, replace its entire akt-cli/ folder with the new bundle so the main instructions and references stay together.

Codex detects newly installed skills automatically. If it does not appear, restart Codex. See OpenAI’s skill documentation for discovery locations and invocation.

Other Coding Agents

Extract the complete akt-cli/ folder into your agent’s supported skills directory, following that agent’s installation instructions. The workflow guidance lives in SKILL.md and references/; agents/openai.yaml supplies optional interface metadata.

From the akt Repository

The skill lives in .agents/skills/akt-cli/ in the akt repository. Copy that complete folder into your project’s or agent’s skills directory, or use the website ZIP above.

Use the Skill

In Codex CLI or its IDE extension, mention $akt-cli explicitly, or let the agent select it when your request matches the skill. Start with a context inspection:

Use $akt-cli to inspect my staging context and list its deployments.

For a deployment preview:

Use $akt-cli to validate deploy.yaml and show the deployment plan for
my staging context. Keep the requested resources and provider constraints.

For troubleshooting, include the real deployment identifier:

Use $akt-cli to inspect deployment <dseq> in my staging context,
check its logs and events, and explain why the service is not ready.

For changes that spend funds or alter a deployment, specify the target context, workload, spending limits, and intended action. The skill guides the agent to preserve those requirements and report the resulting deployment and lease identifiers.

What the Skill Covers

• Context and identity: Inspect configuration, select chain signing or Console managed deployment, and keep credentials out of output.
• Command discovery: Check the installed binary’s version and help before relying on flags or examples.
• Deployment preparation: Write and validate SDL, preserve resource and placement requirements, preview the plan, and choose an explicit bid strategy for unattended work.
• Operations: Use JSON or JSONL output, inspect leases, fetch status, logs, and events, and update or close the intended deployment.
• Recovery: Locate the last successful step, inspect chain and provider state, and continue from existing resources after a partial failure.

The skill works through the CLI with shell access. akt MCP optionally exposes selected Akash operations as tools for MCP clients. For broader deployment, provider, and node-operator guidance, see the separate Akash skill bundle.