The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches data from several official Saudi government domains, including od.data.gov.sa, opendata.cma.org.sa, and www.tdf.gov.sa. These are well-known services for public data access in the region.
[COMMAND_EXECUTION]: The skill utilizes curl and python3 to interact with APIs and process responses. Specifically, it employs python3 -m json.tool for safe formatting of JSON data.
[REMOTE_CODE_EXECUTION]: Automated scanners flagged the pattern of piping curl output directly into python3. Analysis confirms that these commands are used for structural parsing (e.g., extracting resource URLs) and formatting of API responses rather than executing arbitrary remote code or scripts.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted data from external sources and processes it within the agent's execution environment.
Ingestion points: Data is fetched from various API endpoints on the Saudi Open Data Portal and other government sources via curl and Python's urllib library.
Boundary markers: The skill does not implement delimiters or specific instructions to help the agent distinguish between external data and its own system instructions.
Capability inventory: The skill provides access to powerful tools including curl (supporting network requests and file writes via -o) and python3. If a malicious API response provided a crafted URL or payload, it could potentially exploit these tools for server-side request forgery (SSRF) or writing malicious files.
Sanitization: There is no evidence of input validation or output escaping for the data fields (like RESOURCE_URL) retrieved from external APIs before they are used in shell commands.

saudi-opendata