The Agent Skills Directory

[COMMAND_EXECUTION]: The test scripts scripts/test_profile_dataset.py and scripts/test_reconcile_dataset.py execute the skill's own Python utilities using subprocess.check_output and subprocess.run. These calls use sys.executable to target the local scripts within the current environment.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and profile untrusted external data files (CSV, JSONL), creating a surface for indirect prompt injection if malicious instructions are embedded in the data files.
Ingestion points: Untrusted data is ingested by scripts/profile_dataset.py and scripts/reconcile_dataset.py during the profiling and reconciliation phases.
Boundary markers (absent): While the skill uses reporting templates, it lacks explicit delimiter-based isolation for the raw data content in the agent's context.
Capability inventory: The agent is instructed to use a wide range of powerful data manipulation tools (e.g., pandas, Polars, Miller, OpenRefine) to transform data based on the results of the profiling.
Sanitization (absent): The bundled scripts perform type coercion and structural parsing but do not include checks or sanitization to prevent the agent from being influenced by natural language instructions found within data fields.

data-cleaning