The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions for installing the vendor's official SDKs, including @islo-labs/sdk via NPM and the islo package via PyPI. It also references the platform's default container image, ghcr.io/islo-labs/islo-runner:latest.
[COMMAND_EXECUTION]: The documentation describes the islo use <name> -- <command> pattern, which is used to execute arbitrary shell commands within an isolated sandbox. This is a primary functionality of the Islo platform for development and automation.
[INDIRECT_PROMPT_INJECTION]: The skill describes how the agent interacts with external configuration files and repository data, creating a surface for indirect injection.
Ingestion points: Configuration manifests such as islo.yaml and job.toml, as well as source code from external GitHub repositories via the --source flag.
Boundary markers: The platform provides isolation through sandboxes and network-level credential injection (gateway profiles), which act as a security boundary.
Capability inventory: The CLI supports command execution (islo use), file transfers (islo cp), and network operations.
Sanitization: The documentation explicitly instructs users to avoid embedding credentials in source code or configuration, relying instead on the platform's secure gateway for authentication.

platform