The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines a subagent action, Answer_Questions_with_KBAs, which ingests user-provided text through the Input:User_Query variable. This input is interpolated into a prompt template to perform knowledge retrieval and response generation, creating a surface where untrusted data can influence agent behavior.
Ingestion points: Input:User_Query input parameter for the Answer_Questions_with_KBAs action in SKILL.md.
Boundary markers: The subagent instructions include constraints such as "Never provide generic information unless retrieved from knowledge articles," though no technical delimiters or explicit "ignore instructions" warnings for the interpolated data are specified.
Capability inventory: The skill uses the generatePromptResponse capability to search data libraries and generate displayable output for the user.
Sanitization: No specific input sanitization, filtering, or escaping mechanisms are described for the user query before it is processed by the model.
[COMMAND_EXECUTION]: The documentation includes several examples of the sf (Salesforce CLI) utility for project management tasks, such as sf project retrieve start, sf agent validate, sf agent preview, sf agent publish, and sf agent activate. These are standard administrative commands for the Salesforce platform and are used appropriately for the skill's documented purpose of agent configuration and deployment.

connecting-agent-data-library