The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The diagnostic script processes external data from a CSV file, which creates an attack surface for instructions targeting the agent when the output summary is processed.
Ingestion points: scripts/brownian_motion_processes_diagnostics.py reads user-provided data from a CSV file via pd.read_csv and includes column names in its output summary.
Boundary markers: Absent. There are no explicit instructions or delimiters telling the agent to ignore potential instructions embedded in the data or column headers.
Capability inventory: The skill includes file-write capabilities in scripts/brownian_motion_processes_diagnostics.py to save results, but it does not perform network operations or shell command execution.
Sanitization: Present. The script uses pandas.to_numeric with errors="coerce" to process data values, which ensures that non-numeric payloads are neutralized, although column names remain as raw strings.

brownian-motion-processes