The Agent Skills Directory

[SAFE]: The skill provides technical instructions for the Redpanda Connect product. All capabilities described, including network and file system access, are consistent with the intended purpose of the tool and originate from the official vendor.
[COMMAND_EXECUTION]: The instructions cover the standard use of CLI tools rpk and redpanda-connect to run declarative pipelines. This is the primary method of operation for the described software.
[INDIRECT_PROMPT_INJECTION]: The skill documents a data processing pipeline that interacts with untrusted external data. (1) Ingestion points: Data enters the system through various inputs such as http_server and redpanda (Kafka), as described in SKILL.md and references/components.md. (2) Boundary markers: The tool uses a structured YAML configuration to define pipeline stages, providing clear separation between data and logic. (3) Capability inventory: The pipeline engine is capable of writing to files, making HTTP requests, and executing Bloblang transformation logic. (4) Sanitization: Bloblang allows for granular control over message content, enabling users to filter or transform data to remove sensitive information before it reaches output sinks.

connect