The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands using the btp CLI to manage SAP Business Technology Platform resources. These commands are necessary for the skill's primary function of registering and authorizing data destinations within the SAP AI Core environment.\n- [PRIVILEGE_ESCALATION]: The skill instructions include commands for updating service instances (btp update services/instance), which is a privileged operation. This risk is effectively mitigated by Rule 8 in SKILL.md, which mandates a strict verification and confirmation flow: fetching current parameters, displaying a diff to the user, and requiring explicit confirmation before any modification is applied.\n- [INDIRECT_PROMPT_INJECTION]: The predict.py script and associated instructions establish a workflow for ingesting user-provided data for machine learning predictions. Although this data originates from untrusted sources, it is transformed into a structured JSON format and processed by specialized tabular prediction models (sap-rpt-1), minimizing the risk of instructions embedded in the data influencing the agent's behavior.\n- [CREDENTIALS_UNSAFE]: Several scripts (create_data_destination.py, patch_data_destination.py) accept sensitive credentials as command-line arguments to facilitate resource registration. The skill mitigates exposure risk by explicitly instructing the agent never to print or log these values and to rely on the SDK's ability to read from the environment.

genai-hub-tabular-orchestration