git-safety-guard
Fail
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS] (HIGH): The installation instructions direct users to download a binary from the
terraphim/terraphim-aiGitHub repository using theghCLI. This organization is not on the trusted sources list, and the binary itself is not verified for integrity. - [COMMAND_EXECUTION] (MEDIUM): The skill sets up a PreToolUse hook that executes the downloaded binary (
terraphim-agent) and shell scripts to evaluate AI-generated commands. This involves executing arbitrary code downloaded at runtime. - [INDIRECT_PROMPT_INJECTION] (LOW): The guard processes untrusted data (
tool_input.command) through a shell wrapper. While intended to block malicious commands, this creates an attack surface for bypassing filters. - Ingestion points:
tool_input.commandprocessed ingit_safety_guard.sh. - Boundary markers: None present in the wrapper script.
- Capability inventory: Shell execution of
terraphim-agentand feedback loop to the agent. - Sanitization: Relies on internal regex pattern matching within the binary which cannot be statically verified.
Recommendations
- AI detected serious security threats
Audit Metadata