card-onboard
Pass
Audited by Gen Agent Trust Hub on Jul 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the
Bashtool to run a local diagnostic script,scripts/probe.sh, which identifies the state of the system daemon, connection transport, and firmware.\n- [COMMAND_EXECUTION]: Thescripts/probe.shscript performs several local command executions, including process checks (pgrep), network probes to a local daemon (curlto127.0.0.1), and execution of a firmware installation script (install_firmware.py).\n- [PROMPT_INJECTION]: The skill presents an indirect prompt injection attack surface (Category 8) by ingesting untrusted user input and script outputs.\n - Ingestion points: User-provided Wi-Fi SSID and password in
SKILL.md, and JSON data returned byscripts/probe.sh.\n - Boundary markers: Absent; the skill does not use delimiters to wrap untrusted input.\n
- Capability inventory: All subprocess calls, including
pgrep,curl, andpython3, are executed withinscripts/probe.sh, while theBashtool is used inSKILL.md.\n - Sanitization: Absent; user-provided credentials are interpolated into the
/card-wifi-setupcommand inSKILL.mdwithout documented escaping or validation.
Audit Metadata