The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a local Python script scripts/introspect_source.py to automate the discovery of data schemas and relationships.
[EXTERNAL_DOWNLOADS]: The introspection script depends on standard third-party libraries including pandas, pyarrow, and openpyxl from well-known package registries to process different file formats.
[SAFE]: The core functionality of the skill is to read and process local files (Excel, CSV, Parquet) for metadata extraction based on user-defined paths. No unauthorized file access or data exfiltration patterns were detected.
[PROMPT_INJECTION]: The skill instructions are focused on data processing and workflow steps without any attempts to override agent safety protocols or bypass system constraints.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources, presenting an attack surface for indirect prompt injection.
Ingestion points: Data is read from local files and remote databases within scripts/introspect_source.py.
Boundary markers: The skill's output is strictly structured as a data-model.json file, which mitigates the risk of instructions bleeding into the agent context.
Capability inventory: The introspection script is limited to file reading and metadata extraction; it lacks network access or the ability to execute arbitrary shell commands.
Sanitization: The script employs a to_tmdl_name function with regular expressions to ensure that all discovered identifiers are sanitized and safe for use in downstream dashboard generation tools.

data-source-connector