macos-security

Fail

Audited by Gen Agent Trust Hub on Sep 19, 2026

Risk Level: CRITICALDATA_EXFILTRATIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSPRIVILEGE_ESCALATIONPERSISTENCEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill provides explicit instructions and commands for exfiltrating sensitive user data, including Keychains and Safari cookies, to an external server. Evidence: guides/macos-security-deep-dive.md contains a command to upload data to https://attacker.example.com/upload, which is identified as a malicious/phishing URL by automated scanners.\n- [REMOTE_CODE_EXECUTION]: The skill includes instructions to download and execute scripts directly in the shell, which is a high-risk pattern for supply chain attacks. Evidence: payloads.md contains /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)".\n- [EXTERNAL_DOWNLOADS]: The skill downloads binaries and tools from unverified third-party domains. Evidence: payloads.md downloads jtool2 from http://www.newosxbook.com/tools/jtool2.tgz.\n- [PRIVILEGE_ESCALATION]: The skill provides comprehensive instructions on bypassing macOS security controls, including System Integrity Protection (SIP) and Transparency, Consent, and Control (TCC), to gain unauthorized access to protected system data.\n- [PERSISTENCE]: The skill documents numerous methods for establishing persistence on a host, including the abuse of LaunchAgents and LaunchDaemons, which are common tactics for malware to maintain a foothold.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from multiple sources (cookies, web history, notes) without adequate sanitization or boundary markers, and possesses the necessary tool capabilities (network access, file writes) to be exploited via indirect prompt injection.\n
  • Ingestion points: ~/Library/Containers/com.apple.Safari/Data/Library/Cookies/Cookies.binarycookies, ~/Library/Safari/History.db, NoteStore.sqlite.\n
  • Boundary markers: Absent in provided command templates.\n
  • Capability inventory: Shell access (Bash), network operations (curl/wget), file system writes.\n
  • Sanitization: No sanitization or validation of the ingested database content is described.\n- [COMMAND_EXECUTION]: The skill frequently uses administrative privileges (sudo) to execute powerful commands that modify system configurations and access sensitive files.
Recommendations
  • AI detected serious security threats
  • Contains 1 malicious URL(s) - DO NOT USE
Audit Metadata
Risk Level
CRITICAL
Analyzed
Sep 19, 2026, 04:36 PM