The Agent Skills Directory

[COMMAND_EXECUTION]: The cli.sh script executes the nats command-line utility to perform mesh operations.\n- [EXTERNAL_DOWNLOADS]: The skill requires the installation of the nats-py Python package as a dependency.\n- [DATA_EXFILTRATION]: The send_file method in client.py and the file command in cli.sh allow the agent to read local files and transmit them to other agents on the NATS mesh. This is a primary feature of the skill for data sharing and analysis workflows.\n- [REMOTE_CODE_EXECUTION]: An automated scan flagged a command in cli.sh that pipes curl output to python3. Detailed analysis confirms this is a safe operation that uses a hardcoded Python script to parse JSON status information from a local monitoring endpoint rather than executing remote code.\n- [PROMPT_INJECTION]: The skill provides an interface for receiving and processing messages from a mesh network (via serve and subscribe methods), which creates a surface for indirect prompt injection. If an agent processes these messages as instructions, it could be influenced by malicious agents on the same mesh.\n
Ingestion points: Mesh messages received in client.py (serve, subscribe methods) and cli.sh (subscribe command).\n
Boundary markers: None; messages are passed as raw JSON objects to the handling logic.\n
Capability inventory: File system read access (send_file), network communication (NATS protocol), and shell command execution (nats CLI).\n
Sanitization: None; the skill decodes incoming JSON data but does not perform safety validation or sanitization on the payload content.

synapse-client