The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from external database schemas or dbt project files to generate documentation and SQL queries.
Ingestion points: Database metadata such as table names, column names, and constraints are ingested via the sqlalchemy.inspect module. The skill also mentions extracting schema from dbt project files (dbt_project.yml and models/ directory).
Boundary markers: The instructions do not define boundary markers or provide specific directives to the agent to ignore instructions that might be embedded within the database metadata or dbt documentation.
Capability inventory: The skill executes Python code that performs database connections via sqlalchemy, executes SQL queries, and writes various output files to the local filesystem (e.g., data_dictionary.csv, schema_erd.mmd, schema_quick_reference.md).
Sanitization: Metadata retrieved from external sources is interpolated into SQL join patterns and Mermaid ERD diagrams without explicit sanitization or escaping mechanisms.
[COMMAND_EXECUTION]: The skill provides multiple Python code blocks intended for execution by the agent to discover schemas, extract metadata, and generate reports.

schema-mapper