The Agent Skills Directory

[PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data, specifically Amazon product titles and customer reviews, as documented in the workflow within SKILL.md and the templates in references/prompt_templates.md. This ingestion creates a surface for indirect prompt injection, where malicious instructions embedded in product content could potentially influence the agent's behavior.
[COMMAND_EXECUTION]: The scripts/api_client.py file utilizes subprocess.run to execute the system curl utility for communicating with the Sorftime API. While the implementation uses a list of arguments to prevent shell injection, the ability to invoke system binaries remains a sensitive capability.
[EXTERNAL_DOWNLOADS]: The skill performs network operations to mcp.sorftime.com to fetch market reports, product details, and trends. These downloads are core to the skill's functionality and target a well-known service in the Amazon seller tooling ecosystem.
[CREDENTIALS_UNSAFE]: The scripts/api_client.py script programmatically reads the Sorftime API key from the .mcp.json configuration file located in the project root. This is a standard configuration pattern for MCP-based tools but involves reading sensitive local files.

product-research