The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from external files such as JSONL, GML, Parquet, and CSV, which creates a potential surface for indirect prompt injection if malicious instructions are embedded within the imported data.
Ingestion points: Data ingestion occurs through LOAD JSONL, LOAD GML, and LOAD CSV commands as described in importing.md and querying.md.
Boundary markers: The documentation does not specify the use of delimiters or instructions to ignore commands embedded in external data.
Capability inventory: The skill possesses extensive capabilities, including shell command execution (turingdb and turing-parquet CLIs), network communication (via HTTP backends and S3 transfers), and filesystem modification (database writes).
Sanitization: No explicit sanitization or validation steps are provided for the agent to apply to external data before it is processed.
[COMMAND_EXECUTION]: The skill provides instructions for the agent to perform administrative shell operations.
startup.md details installing the turingdb package and using the turingdb CLI to start the database server with various flags (e.g., -demon, -turing-dir).
importing.md describes using the turing-parquet CLI tool to generate graph files from Parquet data.

turingdb