The Agent Skills Directory

[SAFE]: The skill defines a structured methodology for database schema versioning using plain SQL files, promoting transparency and manual review of database changes.
[PROMPT_INJECTION]: The skill includes a utility script that ingests data from an external PostgreSQL database to generate repository files, creating a potential vector for indirect prompt injection.
Ingestion points: references/fetch_missing_objects.py extracts schema definitions, including table/column comments and function bodies, from the PostgreSQL pg_catalog tables.
Boundary markers: The script does not apply specific boundary markers or "ignore instructions" headers to the generated .sql files.
Capability inventory: The script writes to the filesystem (pathlib.Path.write_text) and formats code. The overall skill instructions encourage the agent to manage these files, potentially exposing it to instructions embedded in database metadata.
Sanitization: Content is fetched from the database and written to the repository without semantic validation or sanitization of embedded comments.
[EXTERNAL_DOWNLOADS]: The Python utility utilizes well-known packages (sqlglot, psycopg, questionary, python-dotenv, shandy-sqlfmt) for database connectivity and formatting, which represent standard and reputable development tools.

database-in-source