The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the cargo-ai CLI to perform infrastructure management tasks, including initializing projects (init), planning changes (plan), and deploying resources (deploy). These operations involve shell command execution and interaction with the platform's API.
[EXTERNAL_DOWNLOADS]: The skill downloads the @cargo-ai/cli package from NPM and fetches project cookbooks from the official getcargohq/gtm-skills GitHub repository. These are official vendor resources and well-known services.
[DYNAMIC_EXECUTION]: The skill facilitates the creation of automated workflows and hosted workers. Workflows defined via defineWorkflow are compiled into a server-side execution graph. The framework includes a js() escape hatch for custom logic, which is a core feature of the platform's automation capabilities.
[INDIRECT_PROMPT_INJECTION]: The skill manages workspace state and configuration through local files (.ts definitions and cargo.state.json). It incorporates security controls such as secret() and env() helpers to ensure sensitive credentials remain in the environment and are not committed to state or version control.
Ingestion points: Reads project definitions from local .ts files and tracks deployment state in cargo.state.json (as described in guides/deploy-and-state.md).
Boundary markers: Uses structured define* builders and zod schemas to enforce resource definitions (as described in guides/authoring-resources.md).
Capability inventory: The cargo-ai CLI manages file system access for state tracking and network access for infrastructure reconciliation (as described in references/commands.md).
Sanitization: Workflow logic is validated against a supported JavaScript subset during the build and deployment process (as described in guides/authoring-resources.md).

cargo-project