wifi
Pass
Audited by Gen Agent Trust Hub on Apr 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the
networksetuputility and the macOS internalairportframework utility to manage system-level network configurations. Specifically, it toggles hardware power for theen0interface to disconnect from networks. - [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection via environment data. It captures WiFi SSIDs from the local environment and interpolates them into JSON output without sanitization. A maliciously crafted SSID (e.g., containing escape characters or prompt instructions) could influence the agent's behavior or break the data structure.
- Ingestion points: Output from the
/System/Library/PrivateFrameworks/Apple80211.framework/Versions/Current/Resources/airport -scommand inwifi_cli.sh. - Boundary markers: Absent. SSIDs are directly embedded into a manually constructed JSON string.
- Capability inventory: Execution of shell commands and system-level network configuration changes via
networksetupandairport. - Sanitization: None. The script uses basic
awkandsedformatting which does not escape control characters or potential injection payloads within SSID strings.
Audit Metadata