The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from untrusted external sources, creating a potential vector for indirect prompt injection.
Ingestion points: Data ingestion occurs through read_csv, read_parquet, read_json, and read_postgres_binary functions, as well as remote database queries via postgres_query (found in SKILL.md and references/postgres.md).
Boundary markers: No specific delimiters or safety instructions are provided to the agent to treat the ingested data as untrusted.
Capability inventory: The skill enables significant capabilities, including local file writing (COPY), network communication for remote file access (httpfs extension), and the ability to execute DDL/DML on external PostgreSQL instances (postgres_execute).
Sanitization: There is no evidence of data sanitization or validation routines for the content retrieved from external files or databases.
[COMMAND_EXECUTION]: Instructions are provided for running the DuckDB CLI via the uv tool and executing SQL commands against remote databases (SKILL.md, references/postgres.md).
[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of DuckDB extensions (such as httpfs, spatial, postgres) from the official DuckDB extension repository via the INSTALL command (SKILL.md).

duckdb