type4me-macos-voice-input
Fail
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructions direct the user to clone source code and download binary releases from an external GitHub repository (
github.com/joewongjc/type4me).\n- [REMOTE_CODE_EXECUTION]: The installation process involves executing shell scripts (scripts/build-sherpa.shandscripts/deploy.sh) that are obtained from an external, untrusted source.\n- [PRIVILEGE_ESCALATION]: The skill explicitly instructs the user to bypass macOS security protections by removing the quarantine attribute from the application using thexattr -d com.apple.quarantinecommand.\n- [INDIRECT_PROMPT_INJECTION]:\n - Ingestion points: The skill uses
{selected}text and{clipboard}content as variables in prompt templates for LLM processing.\n - Boundary markers: No delimiters or isolation techniques are mentioned to distinguish untrusted data from the rest of the prompt.\n
- Capability inventory: The skill facilitates network communication with LLM/ASR APIs and possesses text injection capabilities into other applications.\n
- Sanitization: There is no evidence of filtering or sanitization logic applied to the external text before interpolation.
Recommendations
- AI detected serious security threats
Audit Metadata