The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection due to the processing of external database metadata.
Ingestion points: Database schema metadata including table names, column names, and constraints are fetched via SQLAlchemy's inspect module in SKILL.md.
Boundary markers: Absent. The skill instructions do not define delimiters or provide specific warnings to the agent to disregard potential instructions embedded within the database metadata.
Capability inventory: The skill has the capability to write to the local file system using pandas.to_csv() and Python's open().write() methods to create documentation and ERD diagrams as seen in SKILL.md.
Sanitization: Absent. The skill logic directly interpolates metadata into string templates for Markdown files and Mermaid diagrams without sanitizing or escaping the content, relying on the agent's interpretation of the metadata strings.

schema-mapper