commit
Fail
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: HIGHCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION] (HIGH): The skill executes
git commit -m "<message>"via the Bash tool using AI-generated content. This creates a command injection vulnerability if the message contains shell metacharacters like backticks or subshells, as the content is not escaped before being passed to the shell. - [PROMPT_INJECTION] (LOW): The skill lacks defenses against indirect prompt injection. 1. Ingestion points: Data is pulled from
git logandgit diff(SKILL.md). 2. Boundary markers: None. 3. Capability inventory: Use of theBashtool for local command execution. 4. Sanitization: None. The AI-generated output is used directly in a shell command without validation.
Recommendations
- AI detected serious security threats
Audit Metadata