The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates transaction management by executing shell commands using cardano-cli, mkdir, and cat via the Bash tool.
[INDIRECT_PROMPT_INJECTION]: The skill uses shell command templates with placeholders for addresses, amounts, and UTxO hashes, which represents a command injection attack surface.
Ingestion points: Transaction parameters such as <sender-addr>, <recipient>, and <utxo> are typically provided by users or parsed from external data in SKILL.md.
Boundary markers: There are no specific delimiters or instructions for the agent to escape or ignore malicious meta-characters within these inputs.
Capability inventory: The agent has access to a shell via the Bash tool which can execute commands and manipulate the local filesystem.
Sanitization: The instructions do not mandate input validation or sanitization, allowing potentially malicious strings to be interpolated directly into shell commands.
[DYNAMIC_CONTEXT_INJECTION]: The skill metadata contains ! commands that execute cardano-cli version and cardano-cli conway --help at load time. These commands are used to provide the agent with necessary environment details and are considered a benign use of dynamic context injection.

cardano-cli-transactions-operator