The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a Python script (scripts/amazon_keyword_summary.py) to perform its primary function. This script processes JSON-formatted arguments and interacts with the NexScope API.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external Amazon search results via the NexScope API. This data is then presented to the agent, creating a surface where malicious content in product listings could attempt to influence the agent's behavior. The skill mitigates this by instructing the agent to present data in structured tables and use summary fields first.
Ingestion points: Data is fetched via API and loaded from local JSON files saved in the nexscope/ directory.
Boundary markers: The skill provides instructions on how to format the data (tables, specific field labels) which helps maintain context boundaries.
Capability inventory: The skill has the capability to execute the local Python script and read/write to the project's data directory.
Sanitization: The Python script uses standard JSON parsing; the LLM instructions focus on clear presentation logic.
[DATA_EXPOSURE]: The script reads the NEXSCOPE_API_KEY and SESSION_ID from the environment. It persists query results and cache files to the local directory (nexscope/). This is standard behavior for the NexScope platform to maintain session context and reduce API costs.

ecommerce-amazon-keyword-summary