The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill includes hardcoded database credentials (postgres:postgres) as default values in the pull-schemas.sh script and as examples for environment configuration in the migrate.md documentation.
[INDIRECT_PROMPT_INJECTION]: The skill generates snapshots from live database environments that the agent is then instructed to process, creating a surface for potential indirect prompt injection attacks.
Ingestion points: scripts/pull-schemas.sh queries live SQLite and Postgres databases for table names, column definitions, and row counts.
Boundary markers: The generated markdown snapshots lack delimiters or specific instructions to the agent to treat the database-sourced content as untrusted.
Capability inventory: The agent is empowered to execute schema migrations, rollbacks, and code generation via the just utility.
Sanitization: No sanitization or validation of the retrieved metadata is performed before it is presented to the agent in the snapshots.
[COMMAND_EXECUTION]: The skill relies on shell scripts (scripts/pull-schemas.sh) and the just command runner to execute database introspection and migration tools like psql, sqlite3, and goose.

kernel-hominem-database