The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to construct and execute shell commands using curl to download images. In references/url-extract.md, the IMAGE_URL (which is sourced from the url_extraction tool processing an external e-commerce page) is interpolated directly into a shell command. If the extracted data contains shell metacharacters, it could lead to arbitrary command execution.
[EXTERNAL_DOWNLOADS]: The skill is designed to download up to 10 files from arbitrary external domains identified during the URL extraction process. Fetching content from unverified third-party servers is a standard feature for this skill but remains a security consideration.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it processes untrusted data from e-commerce websites.
Ingestion points: Data enters via the url_extraction tool which parses content from user-provided URLs in references/url-extract.md.
Boundary markers: The skill does not define specific delimiters or instructions to ignore embedded commands within the data returned by the extraction tool.
Capability inventory: The skill has access to shell execution (curl, mkdir) and image analysis tools.
Sanitization: There is no logic provided to sanitize or validate the content of the IMAGE_URL or other extracted fields before they are used in shell commands or subsequent prompts.

analyzing-products