The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple shell commands using the GitHub CLI (gh), Git, and Python. These are used to fetch repository metadata, file contents, and diffs, as well as to run helper scripts for schema resolution and URL generation.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of the pyyaml Python package from standard registries to handle dbt configuration files.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources, specifically GitHub Pull Request titles, descriptions, and file diffs. This data is incorporated into the agent's context and output (e.g., notebook titles), creating a surface for potential indirect prompt injection attacks where malicious instructions could be embedded in a PR.
Ingestion points: detailed-guide.md specifies commands to fetch PR titles and file contents from GitHub using gh pr view and gh api.
Boundary markers: No explicit delimiters or boundary markers are documented to wrap untrusted PR content when it is interpolated into the final notebook YAML.
Capability inventory: The skill possesses the ability to execute shell commands (gh, git), read and write local files in /tmp, and generate URLs intended for browser execution.
Sanitization: The resolve_dbt_schema.py script implements a safe_user_path function to prevent directory traversal attacks by validating that paths remain within the current workspace, though it does not explicitly sanitize the natural language metadata extracted from PRs.

monte-carlo-validation-notebook