The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides patterns for processing external data formats (CSV, JSON, SQL) which represents a potential attack surface for indirect prompt injection if the processed data contains instructions intended for the agent.
Ingestion points: The skill uses pd.read_csv, pd.read_parquet, and a custom json_to_dataframe function in SKILL.md to ingest external data.
Boundary markers: The provided code snippets do not include explicit instructions to the agent to ignore instructions embedded within the data content.
Capability inventory: The skill operates in an environment with file system access (Read, Write, Edit, Grep, Glob) and Bash execution capabilities.
Sanitization: The flatten_json and explode_arrays functions focus on structural transformation and do not perform content-based sanitization or filtering of the input data values.

pandas-coder