The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The scripts/generate_schema.py script accepts structured input via the --entities JSON argument and interpolates these values directly into Python code templates without sanitization.
Ingestion points: scripts/generate_schema.py processes the --entities JSON string.
Boundary markers: No boundary markers or "ignore instructions" warnings are used.
Capability inventory: The script creates directories and writes multiple .py files to the filesystem.
Sanitization: No input sanitization or validation is performed on the keys or values within the entities JSON object before they are used in code generation.
[DYNAMIC_EXECUTION]: The skill uses scripts/generate_schema.py to programmatically generate Python source files (models, controllers, schemas) by concatenating strings with user-provided parameters. This pattern of generating executable code from untrusted input poses a risk if the input source is compromised.

database-schema