The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install the Oodle CLI tool using Homebrew or Go from the vendor's official GitHub repository (github.com/oodle-ai/oodle-cli).
[COMMAND_EXECUTION]: The skill uses various command-line tools to manage infrastructure and verify the environment, including kubectl, helm, aws, and gcloud. It adapts setup steps based on the output of these commands.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external data from the Oodle API in the form of 'setup specs' which are used to generate and execute shell commands. It also processes output from infrastructure tools that could potentially be influenced by external actors (e.g., Helm release names or Kubernetes resource metadata).
Ingestion points: Remote JSON/YAML configuration fetched via oodle integrations get-setup-spec and infrastructure metadata from kubectl, helm, and cloud CLIs.
Boundary markers: The instructions include a strict requirement to 'Show the command to be run and wait for user confirmation before executing any command that creates, modifies, or deletes resources.'
Capability inventory: The skill possesses the capability to execute arbitrary shell commands for tool installation and resource management.
Sanitization: The skill contains explicit instructions to redact secrets in all outputs and suggests passing sensitive values through environment variables or temporary files to avoid exposure in command history.

oodle-onboarding