The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructs the agent to execute cat ~/.duckdbrc to verify configuration. This file contains cleartext database credentials, including HOST, USER, and PASSWORD, which are then exposed to the agent's context.
[EXTERNAL_DOWNLOADS]: The skill performs a remote download and installation of the DuckDB CLI from the official domain https://install.duckdb.org using a shell pipe. This is a trusted source but remains a notable execution pattern.
[REMOTE_CODE_EXECUTION]: Employs a curl | sh pattern to install the DuckDB binary directly from a remote URL.
[COMMAND_EXECUTION]: The skill makes extensive use of shell commands to execute SQL queries via duckdb -c, manage local CSV files, and inspect data using head and wc -l.
[PROMPT_INJECTION]: The skill processes untrusted data from an external database, creating a surface for indirect prompt injection.
Ingestion points: Data is queried from StarRocks, exported to local CSV files (e.g., query_result.csv), and then read by the agent using head or cat.
Boundary markers: None identified.
Capability inventory: The agent has the ability to execute arbitrary bash commands and SQL queries, and to present files to the user.
Sanitization: User inputs such as keywords and date ranges are interpolated directly into shell-executed strings without visible escaping or validation.

oci-order-detail-query