The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's workflow involves ingesting business data (supplier names, item descriptions, and category hints) from user-provided JSON files. This data is processed by Python scripts and rendered into markdown reports, which the agent is then instructed to synthesize into a final review. This creates an indirect prompt injection surface where malicious instructions embedded in the input data could influence the agent's synthesis or subsequent decisions.
Ingestion points: The skill reads data from spend.json, pos.json, and suppliers.json via the scripts/spend_categorizer.py, scripts/purchasing_cycle_analyzer.py, and scripts/supplier_consolidation.py scripts.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to ignore potentially malicious embedded content within the processed reports during the synthesis stage.
Capability inventory: The scripts are limited to data processing and markdown generation using the Python standard library. They do not perform network operations or access sensitive system files.
Sanitization: While the scripts use standard JSON parsing, the resulting strings are interpolated directly into markdown f-strings without validation or escaping of the content.

procurement-optimizer