The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to construct HTTP requests using untrusted user-supplied inputs such as BASE_URL, CompanyDB, UserName, and Password (SKILL.md, Quick Start).
Ingestion points: Sensitive environment details and credentials are provided by the user to populate API request templates in SKILL.md and references/service_layer.md.
Boundary markers: The skill does not define delimiters or explicit warnings to prevent the agent from executing instructions potentially embedded within these user-supplied values.
Capability inventory: The skill demonstrates network capabilities by providing curl command examples for API interaction in references/service_layer.md.
Sanitization: There are no instructions for validating, escaping, or filtering the user input before it is interpolated into the HTTP request strings.
[EXTERNAL_DOWNLOADS]: The skill references documentation from SAP's official help portal.
It directs the agent to fetch the "Working with SAP Business One Service Layer" PDF from help.sap.com. As SAP is a well-known technology provider, this reference is documented as a standard practice.
[DATA_EXFILTRATION]: The skill includes technical examples that target a non-whitelisted corporate domain and references application configuration files.
Technical curl examples in references/service_layer.md target the domain sl.bunzlbi.mx.
The skill instructions in SKILL.md include reading and modifying the ServiceLayer/conf/b1s.conf configuration file, which may contain environment-specific settings.

sap-business-one-servicer-layer-sql