The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it ingests and processes external data from CSV files to generate results and dashboards. While the current implementation primarily performs mathematical operations, malicious content in the transaction data could theoretically be used to influence the agent's interpretation of results or downstream tasks.
Ingestion points: The scripts scripts/association_rules.py and scripts/fp_growth.py read from data/transactions.csv and data/category_map.csv.
Boundary markers: No explicit delimiters or instructions are used to treat the CSV content as untrusted data during processing.
Capability inventory: The skill utilizes bash_tool for local script execution and web_search for industry benchmarking.
Sanitization: Data is parsed using standard pandas.read_csv without specific validation or sanitization to filter for natural language instructions embedded in data fields.
[EXTERNAL_DOWNLOADS]: The script scripts/fp_growth.py has a dependency on the mlxtend library for FP-Growth mining. While this is a well-known and reputable package for data science, it represents an external code dependency.
[NO_CODE]: The workflow documentation in SKILL.md references a script named scripts/journey_segmentation.py, but this file is missing from the provided skill contents.

market-basket-analysis