The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it is designed to ingest and process data from untrusted external sources.
Ingestion points: Data is loaded from local files (-T data.csv), AWS S3 buckets (s3://bucket/path/), and Kafka topics (events_topic).
Boundary markers: The instructions lack explicit boundary markers or directives to the agent to disregard any natural language instructions that might be embedded within the processed data files.
Capability inventory: The skill utilizes shell commands (curl) and StarRocks SQL operations (LOAD LABEL, CREATE ROUTINE LOAD, INSERT INTO) which are powerful capabilities that could be targeted if the agent obeys instructions found within the data.
Sanitization: No mechanisms for sanitizing the content of ingested files or validating data against a strict schema to prevent instruction injection are present in the provided templates.
[COMMAND_EXECUTION]: The skill provides functional templates for shell-based commands to facilitate data loading.
Evidence: Several examples in SKILL.md demonstrate the use of curl to interact with the StarRocks Stream Load API (http://fe_host:8030/api/database/table/_stream_load).

starrocks-data-loading