The Agent Skills Directory

[COMMAND_EXECUTION]: The bash script scripts/search_amazon.sh executes system commands including curl to interact with the external API and python3 for data processing.
[EXTERNAL_DOWNLOADS]: The skill retrieves product search data from https://ai-factory.frevana.com which is necessary for its core functionality.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted data from an external API, which could potentially contain malicious instructions targeting the agent.
Ingestion points: API responses retrieved from ai-factory.frevana.com within scripts/search_amazon.sh.
Boundary markers: The instructions in SKILL.md do not implement delimiters or warnings to ignore instructions embedded in the search results.
Capability inventory: The skill includes capabilities for network requests (curl), system command execution (python3), and writing data to local files (cp command to OUTPUT_PATH).
Sanitization: The skill returns the API response body unchanged without performing sanitization or filtering of the content.
[DYNAMIC_EXECUTION]: The bash script utilizes Python heredocs to generate and execute code at runtime for constructing JSON payloads and validating API responses.

amazon-search