Install the packaged Kast skill
The packaged Kast skill is a repository-local directory that tells your
LLM agent how to use Kast. Installing it copies the skill files into your
workspace and writes a .kast-version marker so the same CLI version
skips reinstallation.
Prerequisites
Before you install the skill, you need the Kast CLI installed on your machine. If you haven't done that yet, follow the install guide.
Install the skill
From the workspace root, run the following command to install the skill.
-
Run the install command:
-
Let the command choose the default target directory. It picks from the directories already present in your workspace:
.agents/skills/kast.github/skills/kast.claude/skills/kast
-
Verify the install by checking for the
.kast-versionfile in the target directory. If the same CLI version was already installed, the JSON result showsskipped: true.
Force a reinstall
If you need to replace an existing install, pass --yes=true to skip
the confirmation prompt. If you need a non-default target directory,
pass --target-dir:
What the skill contains
The installed skill directory includes:
SKILL.md— the agent-facing instruction file that describes every available command, its parameters, and expected outputagents/openai.yaml— OpenAI-compatible agent configurationreferences/wrapper-openapi.yaml— OpenAPI specification for thekast skillsubcommand surface
How agents locate the Kast binary
The packaged skill assumes a companion hook sets KAST_CLI_PATH to
an absolute path to the Kast binary before the skill runs. Every
command documented in SKILL.md is then invoked as
"$KAST_CLI_PATH" skill <command> <json>.
Use .github/hooks/resolve-kast-cli-path.sh in your harness to encode
the standard resolution flow (existing KAST_CLI_PATH, then kast on
$PATH, then local build/distribution outputs). For example:
Next steps
- Talk to your agent — how to prompt your agent to use Kast effectively
- Direct CLI usage — when agents call the CLI directly instead of through the skill