The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions in SKILL.md define a high-capability workflow where the agent is expected to execute generated code and manage infrastructure. Step 6 ('Run the App Before Handing Off') explicitly directs the agent to start Docker environments (docker compose up), execute shell scripts (./create-topics.sh), and run Java applications using build tools (./gradlew run or mvn exec:java). While these are necessary for the skill's purpose as a developer tool, they represent significant system capabilities.
[INDIRECT_PROMPT_INJECTION]: The skill generates shell scripts that are vulnerable to command injection through user-supplied data. In scripts/create-topics.sh, the create_topic_local and create_topic_cloud functions construct a command string using the $topic variable and then execute it via eval. If a user provides a topic name containing shell metacharacters (e.g., ; rm -rf /), it would lead to arbitrary command execution when the script is run.
Ingestion points: User-specified topic names and configuration parameters provided during the 'Build Mode' interaction.
Boundary markers: None. The scripts do not escape or validate variables before passing them to the shell's eval function.
Capability inventory: The skill has the capability to write and execute scripts, perform network operations against Kafka clusters, and manage containers.
Sanitization: Absent. The $topic variable is interpolated directly into the command string.

kafka-streams-programming