The Agent Skills Directory

[COMMAND_EXECUTION]: The skill leverages standard command-line tools including gh (GitHub CLI) to fetch pull request metadata and diffs, git to analyze local repository changes, and python3 to execute internal helper scripts for schema resolution and notebook generation.\n- [EXTERNAL_DOWNLOADS]: The skill instructs the agent to install the pyyaml Python library using pip, which is a standard dependency for parsing dbt configuration files.\n- [DYNAMIC_EXECUTION]: Upon generating the validation notebook URL, the generate_notebook_url.py script automatically invokes the system's open command to launch the URL in the default web browser.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes external data from GitHub PRs and local SQL models, establishing a surface for potential indirect prompt injection attacks.\n
Ingestion points: Pull request titles, file paths, diff contents, and SQL model code retrieved via gh api or local git commands.\n
Boundary markers: None. The agent processes the file content directly without specific markers to ignore embedded instructions within the SQL data.\n
Capability inventory: Network access through gh, file read/write operations for script execution, and automated browser interaction through the open command.\n
Sanitization: The included resolve_dbt_schema.py script implements a safe_user_path check to prevent directory traversal attacks, and generate_notebook_url.py performs character sanitization and YAML schema validation before encoding data.

monte-carlo-validation-notebook