The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs shell command execution to facilitate its workflow by using the GitHub CLI (gh) and git to fetch pull request metadata, file contents, and diffs. It also uses subprocess.run to execute the system open command, which launches the browser with a generated URL.
[EXTERNAL_DOWNLOADS]: Fetches configuration files and code content from GitHub repositories using the gh API and CLI tools. These operations target well-known infrastructure as part of the skill's primary function.
[DATA_EXFILTRATION]: Transfers metadata and SQL query logic derived from the local environment or GitHub PRs to the vendor's domain (getmontecarlo.com) via a base64-encoded URL. This is the intended method for importing the generated notebook into the Monte Carlo platform.
[PROMPT_INJECTION]: The skill processes untrusted content from PR diffs and dbt models, presenting an indirect injection surface. * Ingestion points: Reads external data via gh and git commands and direct file reads. * Boundary markers: No explicit delimiter boundaries or instruction-ignore warnings are used when interpolating ingested data. * Capability inventory: The skill can execute shell commands, run Python scripts, and open URLs in the browser. * Sanitization: Includes a sanitize_yaml function to normalize characters before encoding, though it does not semantically validate ingested SQL or YAML content.

generate-validation-notebook