media-osc

Pass

Audited by Gen Agent Trust Hub on Jun 27, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses uv run to execute a local Python script (scripts/oscctl.py) that manages the binary wire format and network communication for the OSC protocol. This is a standard pattern for providing project-specific tooling.
  • [DATA_EXFILTRATION]: The script performs network operations using the socket module to send and receive UDP and TCP packets. It also reads local files via the bundle command to transmit structured payloads. These features are core to the skill's primary purpose as a network controller.
  • [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface through the dump and ping commands, which listen for and display data received from remote network sources.
  • Ingestion points: Network data received via s.recvfrom in the cmd_dump and cmd_ping functions within scripts/oscctl.py.
  • Boundary markers: None. Received data is printed to the agent's context as JSON-formatted text.
  • Capability inventory: The script can read local files and perform outbound network communication.
  • Sanitization: Binary messages are parsed using the struct module; strings are decoded with errors="replace" and then serialized to JSON before being presented to the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 27, 2026, 09:53 AM
Security Audit — agent-trust-hub — media-osc