The Agent Skills Directory

[COMMAND_EXECUTION]: The skill contains Python code that utilizes sqlalchemy and pandas to execute queries against external databases. This includes functions like extract_table_metadata and assess_data_quality which perform data inspection and retrieval.
[COMMAND_EXECUTION]: The skill performs local file system operations by writing generated metadata to Markdown and JSON files (e.g., open(..., 'w')).
[CREDENTIALS_UNSAFE]: The code includes a hardcoded connection string template postgresql://user:pass@host:5432/db. While used as an example, this pattern encourages the use of plain-text credentials in skill code.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the following vectors:
Ingestion points: The extract_table_metadata function reads sample data and schema information directly from a database.
Boundary markers: No delimiters or instructions are present to prevent the agent from interpreting data from the database as instructions.
Capability inventory: The skill possesses SQL execution capabilities via sqlalchemy and file system write access.
Sanitization: Database identifiers (schema, table names) are interpolated into SQL queries using f-strings without escaping or validation, presenting a SQL injection surface.

data-catalog-entry