The Agent Skills Directory

[COMMAND_EXECUTION]: The skill employs several standard Unix CLI tools (such as wc, awk, cut, sort, uniq, and grep) within the /viz ingest mode and the reference/cli_data_inspection.md guide to perform structural analysis and data quality checks on local user files. In the /viz execute mode, the skill generates and executes a self-contained Python plotting script based on user intent and data schema, which is a legitimate use of dynamic execution for the skill's primary purpose.
[PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data files (e.g., CSV, Parquet, JSON) provided by the user. While this introduces an attack surface where data content could potentially influence the agent's generated code or summaries, the risk is mitigated by the skill's structured approach.
Ingestion points: Data is loaded from user-specified file paths during the /viz ingest and /viz explore phases.
Boundary markers: No explicit delimiters or instructions to ignore embedded instructions are implemented for the data processing steps.
Capability inventory: The skill possesses the ability to execute shell commands and generate/run Python code.
Sanitization: There is no explicit sanitization or filtering of the content of the data files before they are processed by the agent.
[EXTERNAL_DOWNLOADS]: The skill's implementation depends on standard, well-known Python packages for data science and visualization, such as numpy, pandas, seaborn, matplotlib, cycler, and palettable.

viz