The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains an attack surface for indirect prompt injection as it processes and displays external data from product marketplaces.
Ingestion points: Untrusted product metadata, such as titles and brand names, is fetched from the NexScope API (which proxies Amazon) in the scripts/image_similarity_finder.py script.
Boundary markers: There are no explicit delimiters or instructions within the scripts or markdown templates to prevent the agent from interpreting instructions that could be maliciously embedded in product titles or descriptions.
Capability inventory: The skill is configured with Bash, Read, and Write tool access, which allows it to execute Python scripts, read data, and write chart images to the local filesystem.
Sanitization: The skill performs basic string truncation for display purposes but does not implement security-focused sanitization to filter out potential injection strings from the retrieved marketplace data.
[EXTERNAL_DOWNLOADS]: The skill's documentation and scripts reference well-known Python libraries, including matplotlib and adjustText, which are required for chart generation and label management.

image-similarity-finder