The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data which could lead to an indirect prompt injection attack.
Ingestion points: The MultiSourceAnalyzer.register_file method in scripts/multi_source_analyzer.py allows the agent to ingest external CSV, Excel, JSON, and Parquet files into the analysis context.
Boundary markers: Absent. The skill instructions do not specify the use of delimiters or 'ignore embedded instructions' warnings when processing or interpolating content from these external data sources.
Capability inventory: The skill possesses significant capabilities including database query execution (execute_sql, query_sql in scripts/toolbox.py), file system writing (save_step_output in scripts/analysis_workflow.py), and network connectivity via the Volcengine API client (scripts/dbw_client.py).
Sanitization: Absent. Data read from external files is loaded directly into DataFrames and may be used to generate further SQL queries or reports without validation of the content for natural language instructions.

database-skill