The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute several shell commands using the amq CLI tool.
Shell commands like amq send, amq drain, and amq thread are used to manage multi-agent communication.
User-supplied inputs, including the <topic> and the design <problem>, are interpolated directly into shell command arguments, which could lead to command injection if not handled by the platform's tool-execution guardrails.
[INDIRECT_PROMPT_INJECTION]: The skill processes messages from external "partner" agents, creating a vector for indirect instructions.
Ingestion points: Data from partner agents is read into the context via amq thread --include-body and amq drain --include-body (found in SKILL.md and references/spec-workflow.md).
Boundary markers: No explicit delimiters or instructions to ignore embedded commands within partner messages are provided.
Capability inventory: The agent has capabilities to execute shell commands, read/write files during the design process, and communicate with other agents via the message queue.
Sanitization: There is no evidence of sanitization for message bodies retrieved from the queue or for user-provided strings used in CLI calls.

amq-spec